… of the GTCMT music informatics group (RSS)
analysis anomaly detection assessment audio audio analysis audio signal processing beat histogram classification data clean-up dataset drum transcription drums education feature extraction features hierarchical classification information machine learning music music genre music genre classification music genre retrieval music information retrieval music performance nmf non-negative matrix factorization onset detection pitch chroma restoration rhythm
2017 |
Wu, Chih-Wei ; Vinton, Mark Blind Bandwidth Extension using K-Means and Support Vector Regression Inproceedings Proceedings of the International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, New Orleans, 2017. Abstract | Links | BibTeX | Tags: audio, audio signal processing, machine learning, quality enhancement, restoration @inproceedings{wu_blind_2017, title = {Blind Bandwidth Extension using K-Means and Support Vector Regression}, author = {Wu, Chih-Wei and Vinton, Mark}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2017/04/Wu-and-Vinton-2017-Blind-Bandwidth-Extension-using-K-Means-and-Suppor.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, publisher = {IEEE}, address = {New Orleans}, abstract = {In this paper, a blind bandwidth extension algorithm for music signals has been proposed. This method applies the K-means algorithm to firstly cluster audio data in the feature space, and constructs multiple envelope predictors for each cluster accordingly using Support Vector Regression (SVR). A set of well-established audio features for Music Information Retrieval (MIR) has been used to characterize the audio content. The resulting system is applied to a variety of music signals without any side information provided. The subjec tive listening test results show that his method can improve the perceptual quality successfully, ut the minor artifacts still leave room for future improvements.}, keywords = {audio, audio signal processing, machine learning, quality enhancement, restoration}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, a blind bandwidth extension algorithm for music signals has been proposed. This method applies the K-means algorithm to firstly cluster audio data in the feature space, and constructs multiple envelope predictors for each cluster accordingly using Support Vector Regression (SVR). A set of well-established audio features for Music Information Retrieval (MIR) has been used to characterize the audio content. The resulting system is applied to a variety of music signals without any side information provided. The subjec tive listening test results show that his method can improve the perceptual quality successfully, ut the minor artifacts still leave room for future improvements. |
Pati, Kumar Ashis ; Lerch, Alexander A Dataset and Method for Electric Guitar Solo Detection in Rock Music Inproceedings Proceedings of the AES Conference on Semantic Audio, Audio Engineering Society (AES), Erlangen, 2017. Abstract | Links | BibTeX | Tags: analysis, audio, dataset, feature analysis, guitar solo detection, machine learning, music, music information retrieval @inproceedings{pati_dataset_2017, title = {A Dataset and Method for Electric Guitar Solo Detection in Rock Music}, author = {Pati, Kumar Ashis and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2017/06/Pati_Lerch_2017_A-Dataset-and-Method-for-Electric-Guitar-Solo-Detection-in-Rock-Music.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the AES Conference on Semantic Audio}, publisher = {Audio Engineering Society (AES)}, address = {Erlangen}, abstract = {This paper explores the problem of automatically detecting electric guitar solos in rock music. A baseline study using standard spectral and temporal audio features in conjunction with an SVM classifier is carried out. To improve detection rates, custom features based on predominant pitch and structural segmentation of songs are designed and investigated. The evaluation of different feature combinations suggests that the combination of all features followed by a post-processing step results in the best accuracy. A macro-accuracy of 78.6% with a solo detection precision of 63.3% is observed for the best feature combination. This publication is accompanied by release of an annotated dataset of electric guitar solos to encourage future research in this area.}, keywords = {analysis, audio, dataset, feature analysis, guitar solo detection, machine learning, music, music information retrieval}, pubstate = {published}, tppubtype = {inproceedings} } This paper explores the problem of automatically detecting electric guitar solos in rock music. A baseline study using standard spectral and temporal audio features in conjunction with an SVM classifier is carried out. To improve detection rates, custom features based on predominant pitch and structural segmentation of songs are designed and investigated. The evaluation of different feature combinations suggests that the combination of all features followed by a post-processing step results in the best accuracy. A macro-accuracy of 78.6% with a solo detection precision of 63.3% is observed for the best feature combination. This publication is accompanied by release of an annotated dataset of electric guitar solos to encourage future research in this area. |
Vidwans, Amruta ; Gururani, Siddharth ; Wu, Chih-Wei ; Subramanian, Vinod ; Swaminathan, Rupak Vignesh ; Lerch, Alexander Objective descriptors for the assessment of student music performances Inproceedings Proceedings of the AES Conference on Semantic Audio, Audio Engineering Society (AES), Erlangen, 2017. Abstract | Links | BibTeX | Tags: analysis, assessment, audio, Computer sound processing, Content analysis (Communication), education, machine learning, music, music performance @inproceedings{vidwans_objective_2017, title = {Objective descriptors for the assessment of student music performances}, author = {Vidwans, Amruta and Gururani, Siddharth and Wu, Chih-Wei and Subramanian, Vinod and Swaminathan, Rupak Vignesh and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2017/06/Vidwans-et-al_2017_Objective-descriptors-for-the-assessment-of-student-music-performances.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the AES Conference on Semantic Audio}, publisher = {Audio Engineering Society (AES)}, address = {Erlangen}, abstract = {Assessment of students’ music performances is a subjective task that requires the judgment of technical correctness as well as aesthetic properties. A computational model automatically evaluating music performance based on objective measurements could ensure consistent and reproducible assessments for, e.g., automatic music tutoring systems. In this study, we investigate the effectiveness of various audio descriptors for assessing performances. Specifically, three different sets of features, including a baseline set, score-independent features, and score-based features, are compared with respect to their efficiency in regression tasks. The results show that human assessments can be modeled to a certain degree, however, the generality of the model still needs further investigation.}, keywords = {analysis, assessment, audio, Computer sound processing, Content analysis (Communication), education, machine learning, music, music performance}, pubstate = {published}, tppubtype = {inproceedings} } Assessment of students’ music performances is a subjective task that requires the judgment of technical correctness as well as aesthetic properties. A computational model automatically evaluating music performance based on objective measurements could ensure consistent and reproducible assessments for, e.g., automatic music tutoring systems. In this study, we investigate the effectiveness of various audio descriptors for assessing performances. Specifically, three different sets of features, including a baseline set, score-independent features, and score-based features, are compared with respect to their efficiency in regression tasks. The results show that human assessments can be modeled to a certain degree, however, the generality of the model still needs further investigation. |
Gururani, Siddharth ; Lerch, Alexander Automatic Sample Detection in Polyphonic Music Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, Suzhou, 2017. Abstract | Links | BibTeX | Tags: @inproceedings{gururani_automatic_2017, title = {Automatic Sample Detection in Polyphonic Music}, author = {Gururani, Siddharth and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2017/07/Gururani_Lerch_2017_Automatic-Sample-Detection-in-Polyphonic-Music.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {Suzhou}, abstract = {The term `sampling' refers to the usage of snippets or loops from existing songs or sample libraries in new songs, mashups, or other music productions. The ability to automatically detect sampling in music is, for instance, beneficial for studies tracking artist influences geographically and temporally. We present a method based on Non-negative Matrix Factorization (NMF) and Dynamic Time Warping (DTW) for the automatic detection of a sample in a pool of songs. The method comprises of two processing steps: first, the DTW alignment path between NMF activations of a song and query sample is computed. Second, features are extracted from this path and used to train a Random Forest classifier to detect the presence of the sample. The method is able to identify samples that are pitch shifted and/or time stretched with approximately 63% F-measure. We evaluate this method against a new publicly available dataset of real-world sample and song pairs.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The term `sampling' refers to the usage of snippets or loops from existing songs or sample libraries in new songs, mashups, or other music productions. The ability to automatically detect sampling in music is, for instance, beneficial for studies tracking artist influences geographically and temporally. We present a method based on Non-negative Matrix Factorization (NMF) and Dynamic Time Warping (DTW) for the automatic detection of a sample in a pool of songs. The method comprises of two processing steps: first, the DTW alignment path between NMF activations of a song and query sample is computed. Second, features are extracted from this path and used to train a Random Forest classifier to detect the presence of the sample. The method is able to identify samples that are pitch shifted and/or time stretched with approximately 63% F-measure. We evaluate this method against a new publicly available dataset of real-world sample and song pairs. |
Wu, Chih-Wei ; Lerch, Alexander Automatic drum transcription using the student-teacher learning paradigm with unlabeled music data Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, Suzhou, 2017. Abstract | Links | BibTeX | Tags: @inproceedings{wu_automatic_2017, title = {Automatic drum transcription using the student-teacher learning paradigm with unlabeled music data}, author = {Wu, Chih-Wei and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2017/07/Wu_Lerch_2017_Automatic-drum-transcription-using-the-student-teacher-learning-paradigm-with.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {Suzhou}, abstract = {Automatic drum transcription is a sub-task of automatic music transcription that converts drum-related audio events into musical notation. While noticeable progress has been made in the past by combining pattern recognition methods with audio signal processing techniques, the major limitation of many state-of-the-art systems still originates from the difficulty of obtaining a meaningful amount of annotated data to support the data-driven algorithms. In this work, we address the challenge of insufficiently labeled data by exploring the possibility of utilizing unlabeled music data from online resources. Specifically, a student neural network is trained using the labels generated from multiple teacher systems. The performance of the model is evaluated on a publicly available dataset. The results show the general viability of using unlabeled music data to improve the performance of drum transcription systems.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Automatic drum transcription is a sub-task of automatic music transcription that converts drum-related audio events into musical notation. While noticeable progress has been made in the past by combining pattern recognition methods with audio signal processing techniques, the major limitation of many state-of-the-art systems still originates from the difficulty of obtaining a meaningful amount of annotated data to support the data-driven algorithms. In this work, we address the challenge of insufficiently labeled data by exploring the possibility of utilizing unlabeled music data from online resources. Specifically, a student neural network is trained using the labels generated from multiple teacher systems. The performance of the model is evaluated on a publicly available dataset. The results show the general viability of using unlabeled music data to improve the performance of drum transcription systems. |
2016 |
Freeman, Jason ; Lerch, Alexander ; Paradis, Matthew (Ed.) Proceedings of the 2nd Web Audio Conference (WAC-2016) Book Atlanta, 2016, ISBN: 978-0-692-61973-5. Links | BibTeX | Tags: audio, proceedings, webaudio @book{freeman_proceedings_2016, title = {Proceedings of the 2nd Web Audio Conference (WAC-2016)}, editor = {Freeman, Jason and Lerch, Alexander and Paradis, Matthew}, url = {https://smartech.gatech.edu/handle/1853/54577}, isbn = {978-0-692-61973-5}, year = {2016}, date = {2016-01-01}, address = {Atlanta}, keywords = {audio, proceedings, webaudio}, pubstate = {published}, tppubtype = {book} } |
Laguna, Christopher ; Lerch, Alexander An Efficient Algorithm For Clipping Detection And Declipping Audio Inproceedings Proceedings of the 141st AES Convention, Audio Engineering Society (AES), Los Angeles, 2016. Abstract | Links | BibTeX | Tags: analysis, audio, audio signal processing, declipping, music, restoration, webaudio @inproceedings{laguna_efficient_2016, title = {An Efficient Algorithm For Clipping Detection And Declipping Audio}, author = {Laguna, Christopher and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/09/Laguna_Lerch_2016_An-Efficient-Algorithm-For-Clipping-Detection-And-Declipping-Audio.pdf}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the 141st AES Convention}, publisher = {Audio Engineering Society (AES)}, address = {Los Angeles}, abstract = {We present an algorithm for end to end declipping, which includes clipping detection and the replacement of clipped samples. To detect regions of clipping, we analyze the signal’s amplitude histogram and the shape of the signal in the time-domain. The sample replacement algorithm uses a two-pass approach: short regions of clipping are replaced in the time-domain and long regions of clipping are replaced in the frequency-domain. The algorithm is robust against different types of clipping and is efficient compared to existing approaches. The algorithm has been implemented in an open source JavaScript client-side web application. Clipping detection is shown to give an f-measure of 0.92 and is robust to the clipping level.}, keywords = {analysis, audio, audio signal processing, declipping, music, restoration, webaudio}, pubstate = {published}, tppubtype = {inproceedings} } We present an algorithm for end to end declipping, which includes clipping detection and the replacement of clipped samples. To detect regions of clipping, we analyze the signal’s amplitude histogram and the shape of the signal in the time-domain. The sample replacement algorithm uses a two-pass approach: short regions of clipping are replaced in the time-domain and long regions of clipping are replaced in the frequency-domain. The algorithm is robust against different types of clipping and is efficient compared to existing approaches. The algorithm has been implemented in an open source JavaScript client-side web application. Clipping detection is shown to give an f-measure of 0.92 and is robust to the clipping level. |
Lu, Yen-Cheng ; Wu, Chih-Wei ; Lu, Chang-Tien ; Lerch, Alexander An Unsupervised Approach to Anomaly Detection in Music Datasets Inproceedings Proceedings of the ACM SIGIR Conference (SIGIR), pp. 749–752, ACM, Pisa, 2016, ISBN: 978-1-4503-4069-4. Abstract | Links | BibTeX | Tags: anomaly detection, audio, data clean-up, dataset, machine learning, music, music genre retrieval, music information retrieval @inproceedings{lu_unsupervised_2016, title = {An Unsupervised Approach to Anomaly Detection in Music Datasets}, author = {Lu, Yen-Cheng and Wu, Chih-Wei and Lu, Chang-Tien and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/07/Lu-et-al_2016_An-Unsupervised-Approach-to-Anomaly-Detection-in-Music-Datasets.pdf}, doi = {10.1145/2911451.2914700}, isbn = {978-1-4503-4069-4}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the ACM SIGIR Conference (SIGIR)}, pages = {749--752}, publisher = {ACM}, address = {Pisa}, series = {SIGIR '16}, abstract = {This paper presents an unsupervised method for systematically identifying anomalies in music datasets. The model integrates categorical regression and robust estimation techniques to infer anomalous scores in music clips. When applied to a music genre recognition dataset, the new method is able to detect corrupted, distorted, or mislabeled audio samples based on commonly used features in music information retrieval. The evaluation results show that the algorithm outperforms other anomaly detection methods and is capable of finding problematic samples identified by human experts. The proposed method introduces a preliminary framework for anomaly detection in music data that can serve as a useful tool to improve data integrity in the future.}, keywords = {anomaly detection, audio, data clean-up, dataset, machine learning, music, music genre retrieval, music information retrieval}, pubstate = {published}, tppubtype = {inproceedings} } This paper presents an unsupervised method for systematically identifying anomalies in music datasets. The model integrates categorical regression and robust estimation techniques to infer anomalous scores in music clips. When applied to a music genre recognition dataset, the new method is able to detect corrupted, distorted, or mislabeled audio samples based on commonly used features in music information retrieval. The evaluation results show that the algorithm outperforms other anomaly detection methods and is capable of finding problematic samples identified by human experts. The proposed method introduces a preliminary framework for anomaly detection in music data that can serve as a useful tool to improve data integrity in the future. |
Lu, Yen-Cheng ; Wu, Chih-Wei ; Lu, Chang-Tien ; Lerch, Alexander Automatic Outlier Detection in Music Genre Datasets Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, New York, 2016. Abstract | Links | BibTeX | Tags: anomaly detection, audio, data clean-up, dataset, music, music genre retrieval, music information retrieval, outlier @inproceedings{lu_automatic_2016, title = {Automatic Outlier Detection in Music Genre Datasets}, author = {Lu, Yen-Cheng and Wu, Chih-Wei and Lu, Chang-Tien and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/07/Lu-et-al_2016_Automatic-Outlier-Detection-in-Music-Genre-Datasets.pdf}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {New York}, series = {ISMIR}, abstract = {Outlier detection, also known as anomaly detection, is an importanttopicthathasbeenstudiedfordecades. Anoutlier detection system is able to identify anomalies in a dataset and thus improve data integrity by removing the detected outliers. It has been successfully applied to different types of data in various fields such as cyber-security, finance, and transportation. In the field of Music Information Re- trieval (MIR), however, the number of related studies is small. In this paper, we introduce different state-of-the-art outlier detection techniques and evaluate their viability in the context of music datasets. More specifically, we present a comparative study of 6 outlier detection algorithms ap- plied to a Music Genre Recognition (MGR) dataset. It is determined how well algorithms can identify mislabeled or corrupted files, and how much the quality of the dataset can be improved. Results indicate that state-of-the-art anomaly detection systems have problems identifying anomalies in MGR datasets reliably.}, keywords = {anomaly detection, audio, data clean-up, dataset, music, music genre retrieval, music information retrieval, outlier}, pubstate = {published}, tppubtype = {inproceedings} } Outlier detection, also known as anomaly detection, is an importanttopicthathasbeenstudiedfordecades. Anoutlier detection system is able to identify anomalies in a dataset and thus improve data integrity by removing the detected outliers. It has been successfully applied to different types of data in various fields such as cyber-security, finance, and transportation. In the field of Music Information Re- trieval (MIR), however, the number of related studies is small. In this paper, we introduce different state-of-the-art outlier detection techniques and evaluate their viability in the context of music datasets. More specifically, we present a comparative study of 6 outlier detection algorithms ap- plied to a Music Genre Recognition (MGR) dataset. It is determined how well algorithms can identify mislabeled or corrupted files, and how much the quality of the dataset can be improved. Results indicate that state-of-the-art anomaly detection systems have problems identifying anomalies in MGR datasets reliably. |
Winters, Michael R; Gururani, Siddharth ; Lerch, Alexander Automatic Practice Logging: Introduction, Dataset & Preliminary Study Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), New York, 2016. Abstract | Links | BibTeX | Tags: alignment, analysis, audio, audio analysis, music, music information retrieval, music performance, piano, practice @inproceedings{winters_automatic_2016, title = {Automatic Practice Logging: Introduction, Dataset & Preliminary Study}, author = {Winters, R Michael and Gururani, Siddharth and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/07/Winters-et-al_2016_Automatic-Practice-Logging.pdf}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, address = {New York}, abstract = {Musicians spend countless hours practicing their instru- ments. To document and organize this time, musicians com- monly use practice charts to log their practice. However, manual techniques require time, dedication, and experience to master, are prone to fallacy and omission, and ultimately can not describe the subtle variations in each repetition. This paper presents an alternative: by analyzing and clas- sifying the audio recorded while practicing, logging could occur automatically, with levels of detail, accuracy, and ease that would not be possible otherwise. Towards this goal, we introduce the problem of Automatic Practice Logging (APL), including a discussion of the benefits and unique challenges it raises. We then describe a new dataset of over 600 annotated recordings of solo piano practice, which can be used to design and evaluate APL systems. After fram- ing our approach to the problem, we present an algorithm designed to align short segments of practice audio with reference recordings using pitch chroma and dynamic time warping.}, keywords = {alignment, analysis, audio, audio analysis, music, music information retrieval, music performance, piano, practice}, pubstate = {published}, tppubtype = {inproceedings} } Musicians spend countless hours practicing their instru- ments. To document and organize this time, musicians com- monly use practice charts to log their practice. However, manual techniques require time, dedication, and experience to master, are prone to fallacy and omission, and ultimately can not describe the subtle variations in each repetition. This paper presents an alternative: by analyzing and clas- sifying the audio recorded while practicing, logging could occur automatically, with levels of detail, accuracy, and ease that would not be possible otherwise. Towards this goal, we introduce the problem of Automatic Practice Logging (APL), including a discussion of the benefits and unique challenges it raises. We then describe a new dataset of over 600 annotated recordings of solo piano practice, which can be used to design and evaluate APL systems. After fram- ing our approach to the problem, we present an algorithm designed to align short segments of practice audio with reference recordings using pitch chroma and dynamic time warping. |
Wu, Chih-Wei ; Gururani, Siddharth ; Laguna, Christopher ; Pati, Ashis ; Vidwans, Amruta ; Lerch, Alexander Towards the Objective Assessment of Music Performances Inproceedings Proceedings of the International Conference on Music Perception and Cognition (ICMPC), pp. 99–103, San Francisco, 2016, ISBN: 1-879346-65-5. Abstract | Links | BibTeX | Tags: analysis, assessment, audio, education, machine learning, music, music performance @inproceedings{wu_towards_2016, title = {Towards the Objective Assessment of Music Performances}, author = {Wu, Chih-Wei and Gururani, Siddharth and Laguna, Christopher and Pati, Ashis and Vidwans, Amruta and Lerch, Alexander}, url = {http://www.icmpc.org/icmpc14/proceedings.html}, isbn = {1-879346-65-5}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the International Conference on Music Perception and Cognition (ICMPC)}, pages = {99--103}, address = {San Francisco}, abstract = {The qualitative assessment of music performances is a task that is influenced by technical correctness, deviations from established performance standards, and aesthetic judgment. Despite its inherently subjective nature, a quantitative overall assessment is often desired, as exemplified by US all-state auditions or other competitions. A model that automatically generates assessments from the audio data would allow for objective assessments and enable musically intelligent computer-assisted practice sessions for students learning an instrument. While existing systems are already able to provide similar basic functionality, they rely on the musical score as prior knowledge. In this paper, we present a score-independent system for assessing student instrument performances based on audio recordings. This system aims to characterize the performance with both well-established and custom-designed audio features, model expert assessments of student performances, and predict the assessment of unknown audio recordings. The results imply the viability of modeling human assessment with score-independent audio features. Results could lead towards more general software music tutoring systems that do not require score information for the assessment of student music performances.}, keywords = {analysis, assessment, audio, education, machine learning, music, music performance}, pubstate = {published}, tppubtype = {inproceedings} } The qualitative assessment of music performances is a task that is influenced by technical correctness, deviations from established performance standards, and aesthetic judgment. Despite its inherently subjective nature, a quantitative overall assessment is often desired, as exemplified by US all-state auditions or other competitions. A model that automatically generates assessments from the audio data would allow for objective assessments and enable musically intelligent computer-assisted practice sessions for students learning an instrument. While existing systems are already able to provide similar basic functionality, they rely on the musical score as prior knowledge. In this paper, we present a score-independent system for assessing student instrument performances based on audio recordings. This system aims to characterize the performance with both well-established and custom-designed audio features, model expert assessments of student performances, and predict the assessment of unknown audio recordings. The results imply the viability of modeling human assessment with score-independent audio features. Results could lead towards more general software music tutoring systems that do not require score information for the assessment of student music performances. |
Wu, Chih-Wei ; Lerch, Alexander On Drum Playing Technique Detection in Polyphonic Mixtures Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), New York, 2016. Abstract | Links | BibTeX | Tags: audio, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization, playing technique @inproceedings{wu_drum_2016, title = {On Drum Playing Technique Detection in Polyphonic Mixtures}, author = {Wu, Chih-Wei and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/07/Wu_Lerch_2016_On-Drum-Playing-Technique-Detection-in-Polyphonic-Mixtures.pdf}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, address = {New York}, abstract = {In this paper, the problem of drum playing technique detection in polyphonic mixtures of music is addressed. We focus on the identification of 4 rudimentary techniques: strike, buzz roll, flam, and drag. The specifics and the challenges of this task are being discussed, and different sets of features are compared, including various features extracted from NMF-based activation functions, as well as baseline spectral features. We investigate the capabil- ities and limitations of the presented system in the case of real-world recordings and polyphonic mixtures. To de- sign and evaluate the system, two datasets are introduced: a training dataset generated from individual drum hits, and ad- ditional annotations of the well-known ENST drum dataset minus one subset as test dataset. The results demonstrate issues with the traditionally used spectral features, and in- dicate the potential of using NMF activation functions for playing technique detection, however, the performance of polyphonic music still leaves room for future improvement.}, keywords = {audio, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization, playing technique}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, the problem of drum playing technique detection in polyphonic mixtures of music is addressed. We focus on the identification of 4 rudimentary techniques: strike, buzz roll, flam, and drag. The specifics and the challenges of this task are being discussed, and different sets of features are compared, including various features extracted from NMF-based activation functions, as well as baseline spectral features. We investigate the capabil- ities and limitations of the presented system in the case of real-world recordings and polyphonic mixtures. To de- sign and evaluate the system, two datasets are introduced: a training dataset generated from individual drum hits, and ad- ditional annotations of the well-known ENST drum dataset minus one subset as test dataset. The results demonstrate issues with the traditionally used spectral features, and in- dicate the potential of using NMF activation functions for playing technique detection, however, the performance of polyphonic music still leaves room for future improvement. |
Xambo, Anna ; Lerch, Alexander ; Freeman, Jason Learning to code through MIR Inproceedings Late Breaking Demo (Extended Abstract), Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), New York, 2016. Abstract | Links | BibTeX | Tags: earsketch, education, learning, music information retrieval @inproceedings{xambo_learning_2016, title = {Learning to code through MIR}, author = {Xambo, Anna and Lerch, Alexander and Freeman, Jason}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/08/Xambo-et-al.-2016-Learning-to-code-through-MIR.pdf}, year = {2016}, date = {2016-01-01}, booktitle = {Late Breaking Demo (Extended Abstract), Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, address = {New York}, abstract = {An approach to teaching computer science (CS) in high- schools is using EarSketch, a free online tool for teaching CS concepts while making music. In this demonstration we present the potential of teaching music information retrieval (MIR) concepts using EarSketch. The aim is twofold: to discuss the benefits of introducing MIR concepts in the classroom and to shed light on how MIR concepts can be gently introduced in a CS curriculum. We conclude by identifying the advantagesofteachingMIR inthe classroom and pointing to future directions for research.}, keywords = {earsketch, education, learning, music information retrieval}, pubstate = {published}, tppubtype = {inproceedings} } An approach to teaching computer science (CS) in high- schools is using EarSketch, a free online tool for teaching CS concepts while making music. In this demonstration we present the potential of teaching music information retrieval (MIR) concepts using EarSketch. The aim is twofold: to discuss the benefits of introducing MIR concepts in the classroom and to shed light on how MIR concepts can be gently introduced in a CS curriculum. We conclude by identifying the advantagesofteachingMIR inthe classroom and pointing to future directions for research. |
2015 |
Gupta, Udit ; Moore II, Elliot ; Lerch, Alexander On the Perceptual Relevance of Objective Source Separation Measures for Singing Voice Separation Inproceedings Proceedings of the Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), IEEE, New Paltz, 2015. Abstract | Links | BibTeX | Tags: assessment, audio, listening test, machine learning, music, singing voice separation @inproceedings{gupta_perceptual_2015, title = {On the Perceptual Relevance of Objective Source Separation Measures for Singing Voice Separation}, author = {Gupta, Udit and Moore II, Elliot and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/10/Gupta-et-al_2015_On-the-Perceptual-Relevance-of-Objective-Source-Separation-Measures-for-Singing.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)}, publisher = {IEEE}, address = {New Paltz}, abstract = {Singing Voice Separation (SVS) is a task which uses audio source separation methods to isolate the vocal component from the back- ground accompaniment for a song mix. This paper discusses the methods of evaluating SVS algorithms, and determines how the current state of the art measures correlate to human perception. A modified ITU-R BS.1543 MUSHRA test is used to get the human perceptual ratings for the outputs of various SVS algorithms, which are correlated with widely used objective measures for source sep- aration quality. The results show that while the objective measures provide a moderate correlation with perceived intelligibility and isolation, they may not adequately assess the overall perceptual quality.}, keywords = {assessment, audio, listening test, machine learning, music, singing voice separation}, pubstate = {published}, tppubtype = {inproceedings} } Singing Voice Separation (SVS) is a task which uses audio source separation methods to isolate the vocal component from the back- ground accompaniment for a song mix. This paper discusses the methods of evaluating SVS algorithms, and determines how the current state of the art measures correlate to human perception. A modified ITU-R BS.1543 MUSHRA test is used to get the human perceptual ratings for the outputs of various SVS algorithms, which are correlated with widely used objective measures for source sep- aration quality. The results show that while the objective measures provide a moderate correlation with perceived intelligibility and isolation, they may not adequately assess the overall perceptual quality. |
Lykartsis, Athanasios ; Lerch, Alexander ; Weinzierl, Stefan Analysis of Speech Rhythm for Language Identification Based on Beat Histograms Inproceedings Proceedings of the DAGA (Jahrestagung fur Akustik), Nuremberg, 2015. Links | BibTeX | Tags: audio, beat histogram, language, machine learning, rhythm, speech @inproceedings{lykartsis_analysis_2015, title = {Analysis of Speech Rhythm for Language Identification Based on Beat Histograms}, author = {Lykartsis, Athanasios and Lerch, Alexander and Weinzierl, Stefan}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/06/Lykartsis%20et%20al_2015_Analysis%20of%20Speech%20Rhythm%20for%20Language%20Identification%20Based%20on%20Beat%20Histograms.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the DAGA (Jahrestagung fur Akustik)}, address = {Nuremberg}, keywords = {audio, beat histogram, language, machine learning, rhythm, speech}, pubstate = {published}, tppubtype = {inproceedings} } |
Lykartsis, Athanasios ; Lerch, Alexander Beat Histogram Features for Rythm-based Musical Genre Classification Using Multiple Novelty Functions Inproceedings Proceedings of the International Conference on Digital Audio Effects (DAFX), Trondheim, Norway, 2015. Abstract | Links | BibTeX | Tags: audio, audio analysis, beat histogram, genre, machine learning, music, music genre classification, music information retrieval, rhythm @inproceedings{lykartsis_beat_2015, title = {Beat Histogram Features for Rythm-based Musical Genre Classification Using Multiple Novelty Functions}, author = {Lykartsis, Athanasios and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/12/DAFx-15_submission_42-1.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the International Conference on Digital Audio Effects (DAFX)}, address = {Trondheim, Norway}, abstract = {In this paper we present beat histogram features for multiple level rhythmdescriptionandevaluatetheminamusicalgenreclassifica- tion task. Audio features pertaining to various musical content cat- egories and their related novelty functions are extracted as a basis for the creation of beat histograms. The proposed features capture not only amplitude, but also tonal and general spectral changes in the signal, aiming to represent as much rhythmic information as possible. The most and least informative features are identi- fied through feature selection methods and are then tested using Support Vector Machines on five genre datasets concerning classi- fication accuracy against a baseline feature set. Results show that the presented features provide comparable classification accuracy with respect to other genre classification approaches using period- icity histograms and display a performance close to that of much more elaborate up-to-date approaches for rhythm description. The use of bar boundary annotations for the texture frames has pro- vided an improvement for the dance-oriented Ballroom dataset. The comparably small number of descriptors and the possibility of evaluating the influence of specific signal components to the gen- eral rhythmic content encourage the further use of the method in rhythm description tasks.}, keywords = {audio, audio analysis, beat histogram, genre, machine learning, music, music genre classification, music information retrieval, rhythm}, pubstate = {published}, tppubtype = {inproceedings} } In this paper we present beat histogram features for multiple level rhythmdescriptionandevaluatetheminamusicalgenreclassifica- tion task. Audio features pertaining to various musical content cat- egories and their related novelty functions are extracted as a basis for the creation of beat histograms. The proposed features capture not only amplitude, but also tonal and general spectral changes in the signal, aiming to represent as much rhythmic information as possible. The most and least informative features are identi- fied through feature selection methods and are then tested using Support Vector Machines on five genre datasets concerning classi- fication accuracy against a baseline feature set. Results show that the presented features provide comparable classification accuracy with respect to other genre classification approaches using period- icity histograms and display a performance close to that of much more elaborate up-to-date approaches for rhythm description. The use of bar boundary annotations for the texture frames has pro- vided an improvement for the dance-oriented Ballroom dataset. The comparably small number of descriptors and the possibility of evaluating the influence of specific signal components to the gen- eral rhythmic content encourage the further use of the method in rhythm description tasks. |
Lykartsis, Athanasios ; Wu, Chih-Wei ; Lerch, Alexander Beat Histogram Features from NMF-Based Novelty Functions for Music Classification Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, Malaga, 2015. Abstract | Links | BibTeX | Tags: audio, audio analysis, beat histogram, machine learning, music, music genre classification, music information retrieval, nmf, non-negative matrix factorization, novelty function @inproceedings{lykartsis_beat_2015-1, title = {Beat Histogram Features from NMF-Based Novelty Functions for Music Classification}, author = {Lykartsis, Athanasios and Wu, Chih-Wei and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/10/Lykartsis-et-al_2015_Beat-Histogram-Features-from-NMF-Based-Novelty-Functions-for-Music.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {Malaga}, abstract = {In this paper we present novel rhythm features derived from drum tracks extracted from polyphonic music and evaluate them in a genre classification task. Musical excerpts are analyzed using an optimized, partially fixed Non-Negative Matrix Factorization (NMF) method and beat histogram features are calculated on basis of the resulting activation functions for each one out of three drum tracks extracted (Hi-Hat, SnareDrumandBassDrum). Thefeaturesareeval- uated on two widely used genre datasets (GTZAN and Ball- room) using standard classification methods, concerning the achieved overall classification accuracy. Furthermore, their suitability in distinguishing between rhythmically sim- ilar genres and the performance of the features resulting from individual activation functions is discussed. Results show that the presented NMF-based beat histogram features can provide comparable performance to other classification systems, while considering strictly drum patterns.}, keywords = {audio, audio analysis, beat histogram, machine learning, music, music genre classification, music information retrieval, nmf, non-negative matrix factorization, novelty function}, pubstate = {published}, tppubtype = {inproceedings} } In this paper we present novel rhythm features derived from drum tracks extracted from polyphonic music and evaluate them in a genre classification task. Musical excerpts are analyzed using an optimized, partially fixed Non-Negative Matrix Factorization (NMF) method and beat histogram features are calculated on basis of the resulting activation functions for each one out of three drum tracks extracted (Hi-Hat, SnareDrumandBassDrum). Thefeaturesareeval- uated on two widely used genre datasets (GTZAN and Ball- room) using standard classification methods, concerning the achieved overall classification accuracy. Furthermore, their suitability in distinguishing between rhythmically sim- ilar genres and the performance of the features resulting from individual activation functions is discussed. Results show that the presented NMF-based beat histogram features can provide comparable performance to other classification systems, while considering strictly drum patterns. |
O'Brien, Cian ; Lerch, Alexander Genre-Specific Key Profiles Inproceedings Proceedings of the International Computer Music Conference (ICMC), ICMA, Denton, 2015. Abstract | Links | BibTeX | Tags: audio, audio analysis, key profile, machine learning, music, music genre, music information retrieval, musical key, pitch chroma, pitch content @inproceedings{obrien_genre-specific_2015, title = {Genre-Specific Key Profiles}, author = {O'Brien, Cian and Lerch, Alexander}, url = {"http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/09/O'Brien_Lerch_2015_Genre-Specific%20Key%20Profiles.pdf"}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the International Computer Music Conference (ICMC)}, publisher = {ICMA}, address = {Denton}, abstract = {The most common approaches to the automatic recognition of musical key are template-based, i.e., an extracted pitch chroma vector is compared to a template key profile in order to identify the most similar key. General as well as domain- specific templates have been used in the past, but to the au- thors best knowledge there has been no study that evaluated genre-specific key profiles extracted from the audio signal. We investigate the pitch chroma distributions for 9 different gen- res, their distances, and the degree to which these genres can be identified using these distributions when utilizing different strategies for achieving key-invariance.}, keywords = {audio, audio analysis, key profile, machine learning, music, music genre, music information retrieval, musical key, pitch chroma, pitch content}, pubstate = {published}, tppubtype = {inproceedings} } The most common approaches to the automatic recognition of musical key are template-based, i.e., an extracted pitch chroma vector is compared to a template key profile in order to identify the most similar key. General as well as domain- specific templates have been used in the past, but to the au- thors best knowledge there has been no study that evaluated genre-specific key profiles extracted from the audio signal. We investigate the pitch chroma distributions for 9 different gen- res, their distances, and the degree to which these genres can be identified using these distributions when utilizing different strategies for achieving key-invariance. |
Wu, Chih-Wei ; Lerch, Alexander Drum Transcription using Partially Fixed Non-Negative Matrix Factorization Inproceedings Proceedings of the European Signal Processing Conference (EUSIPCO), EURASIP, Nice, 2015. Abstract | Links | BibTeX | Tags: audio, audio analysis, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization @inproceedings{wu_drum_2015, title = {Drum Transcription using Partially Fixed Non-Negative Matrix Factorization}, author = {Wu, Chih-Wei and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/09/Wu_Lerch_2015_Drum%20Transcription%20using%20Partially%20Fixed%20Non-Negative%20Matrix%20Factorization.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the European Signal Processing Conference (EUSIPCO)}, publisher = {EURASIP}, address = {Nice}, abstract = {In this paper, a drum transcription algorithm using partially fixed non-negative matrix factorization is presented. The pro- posed method allows users to identify percussive events in complex mixtures with a minimal training set. The algorithm decomposes the music signal into two parts: percussive part with pre-defined drum templates and harmonic part with un- defined entries. The harmonic part is able to adapt to the music content, allowing the algorithm to work in polyphonic mixtures. Drum event times can be simply picked from the percussive activation matrix with onset detection. The system is efficient and robust even with a minimal training set. The recognition rates for the ENST dataset vary from 56.7 to 78.9% for three percussive instruments extracted from polyphonic music.}, keywords = {audio, audio analysis, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, a drum transcription algorithm using partially fixed non-negative matrix factorization is presented. The pro- posed method allows users to identify percussive events in complex mixtures with a minimal training set. The algorithm decomposes the music signal into two parts: percussive part with pre-defined drum templates and harmonic part with un- defined entries. The harmonic part is able to adapt to the music content, allowing the algorithm to work in polyphonic mixtures. Drum event times can be simply picked from the percussive activation matrix with onset detection. The system is efficient and robust even with a minimal training set. The recognition rates for the ENST dataset vary from 56.7 to 78.9% for three percussive instruments extracted from polyphonic music. |
Wu, Chih-Wei ; Lerch, Alexander Drum Transcription using Partially Fixed Non-Negative Matrix Factorization With Template Adaptation Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, Malaga, 2015. Abstract | Links | BibTeX | Tags: adaptation, audio, audio analysis, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization @inproceedings{wu_drum_2015-1, title = {Drum Transcription using Partially Fixed Non-Negative Matrix Factorization With Template Adaptation}, author = {Wu, Chih-Wei and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/10/Wu_Lerch_2015_Drum-Transcription-using-Partially-Fixed-Non-Negative-Matrix-Factorization-With.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {Malaga}, abstract = {In this paper, a template adaptive drum transcription algo- rithm using partially fixed Non-negative Matrix Factoriza- tion (NMF) is presented. The proposed method detects per- cussive events in complex mixtures of music with a minimal training set. The algorithm decomposes the music signal into two dictionaries: a percussive dictionary initialized with pre-defined drum templates and a harmonic dictionary initialized with undefined entries. The harmonic dictionary is adapted to the non-percussive music content in a standard NMF procedure. The percussive dictionary is adapted to each individual signal in an iterative scheme: it is fixed during the decomposition process, and is updated based on the result of the previous convergence. Two template adap- tation methods are proposed to provide more flexibility and robustness in the case of unknown data. The performance of the proposed system has been evaluated and compared to state of the art systems. The results show that template adaptation improves the transcription performance, and the detection accuracy is in the same range as more complex systems.}, keywords = {adaptation, audio, audio analysis, drum transcription, drums, machine learning, music, music information retrieval, nmf, non-negative matrix factorization}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, a template adaptive drum transcription algo- rithm using partially fixed Non-negative Matrix Factoriza- tion (NMF) is presented. The proposed method detects per- cussive events in complex mixtures of music with a minimal training set. The algorithm decomposes the music signal into two dictionaries: a percussive dictionary initialized with pre-defined drum templates and a harmonic dictionary initialized with undefined entries. The harmonic dictionary is adapted to the non-percussive music content in a standard NMF procedure. The percussive dictionary is adapted to each individual signal in an iterative scheme: it is fixed during the decomposition process, and is updated based on the result of the previous convergence. Two template adap- tation methods are proposed to provide more flexibility and robustness in the case of unknown data. The performance of the proposed system has been evaluated and compared to state of the art systems. The results show that template adaptation improves the transcription performance, and the detection accuracy is in the same range as more complex systems. |
Zhou, Xinquan ; Lerch, Alexander Chord Detection Using Deep Learning Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), ISMIR, Malaga, 2015. Abstract | Links | BibTeX | Tags: audio, audio analysis, chord detection, deep learning, machine learning, music, music information retrieval @inproceedings{zhou_chord_2015, title = {Chord Detection Using Deep Learning}, author = {Zhou, Xinquan and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/10/Zhou_Lerch_2015_Chord-Detection-Using-Deep-Learning.pdf}, year = {2015}, date = {2015-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, publisher = {ISMIR}, address = {Malaga}, abstract = {In this paper, we utilize deep learning to learn high-level features for audio chord detection. The learned features, obtained by a deep network in bottleneck architecture, give promising results and outperform state-of-the-art systems. We present and evaluate the results for various methods and configurations, including input pre-processing, a bottleneck architecture, and SVMs vs. HMMs for chord classification.}, keywords = {audio, audio analysis, chord detection, deep learning, machine learning, music, music information retrieval}, pubstate = {published}, tppubtype = {inproceedings} } In this paper, we utilize deep learning to learn high-level features for audio chord detection. The learned features, obtained by a deep network in bottleneck architecture, give promising results and outperform state-of-the-art systems. We present and evaluate the results for various methods and configurations, including input pre-processing, a bottleneck architecture, and SVMs vs. HMMs for chord classification. |
2014 |
von Coler, Henrik ; Lerch, Alexander CMMSD: A Data Set for Note-Level Segmentation of Monophonic Music Inproceedings Proceedings of the AES 53rd International Conference on Semantic Audio, Audio Engineering Society (AES), London, UK, 2014. Abstract | Links | BibTeX | Tags: audio, audio analysis, dataset, machine learning, monophonic, music, music information retrieval, onset detection @inproceedings{coler_cmmsd:_2014, title = {CMMSD: A Data Set for Note-Level Segmentation of Monophonic Music}, author = {Coler, Henrik von and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/04/Coler_Lerch_2014_CMMSD.pdf}, year = {2014}, date = {2014-01-01}, booktitle = {Proceedings of the AES 53rd International Conference on Semantic Audio}, publisher = {Audio Engineering Society (AES)}, address = {London, UK}, abstract = {A musical data set for note-level segmentation of monophonic music is presented. It contains 36 excerpts from commercial recordings of monophonic classical western music and features the instrument groups strings, woodwind and brass. The excerpts are self-contained phrases with a mean length of 17.97 seconds and an average of 20 notes. All phrases are played in moderate tempo, mostly with significant amounts of expressive articulation. A manually annotated ground truth splits each item into a sequence of the three states note, transition and rest. The set is designed as an open source project, aiming at the development and evaluation of algorithms for segmentation, music performance analysis and feature selection. This paper presents the process of ground truth labeling and a detailed description of the data set and its properties.}, keywords = {audio, audio analysis, dataset, machine learning, monophonic, music, music information retrieval, onset detection}, pubstate = {published}, tppubtype = {inproceedings} } A musical data set for note-level segmentation of monophonic music is presented. It contains 36 excerpts from commercial recordings of monophonic classical western music and features the instrument groups strings, woodwind and brass. The excerpts are self-contained phrases with a mean length of 17.97 seconds and an average of 20 notes. All phrases are played in moderate tempo, mostly with significant amounts of expressive articulation. A manually annotated ground truth splits each item into a sequence of the three states note, transition and rest. The set is designed as an open source project, aiming at the development and evaluation of algorithms for segmentation, music performance analysis and feature selection. This paper presents the process of ground truth labeling and a detailed description of the data set and its properties. |
Lerch, Alexander Music Information Retrieval Incollection Weinzierl, Stefan (Ed.): Akustische Grundlagen der Musik, (5), pp. 79–102, Laaber, 2014, ISBN: 978-3-89007-699-7. BibTeX | Tags: music information retrieval @incollection{lerch_music_2014, title = {Music Information Retrieval}, author = {Lerch, Alexander}, editor = {Weinzierl, Stefan}, isbn = {978-3-89007-699-7}, year = {2014}, date = {2014-01-01}, booktitle = {Akustische Grundlagen der Musik}, number = {5}, pages = {79--102}, publisher = {Laaber}, series = {Handbuch der Systematischen Musikwissenschaft}, keywords = {music information retrieval}, pubstate = {published}, tppubtype = {incollection} } |
2013 |
Kraft, Sebastian ; Lerch, Alexander ; Zölzer, Udo The Tonalness Spectrum: Feature-Based Estimation of Tonal Components Inproceedings Proceedings of the 16th International Conference on Digital Audio Effects, Maynooth, 2013. Abstract | Links | BibTeX | Tags: audio, audio analysis, features, machine learning, music, music information retrieval, tonality, tonalness @inproceedings{kraft_tonalness_2013, title = {The Tonalness Spectrum: Feature-Based Estimation of Tonal Components}, author = {Kraft, Sebastian and Lerch, Alexander and Z\"{o}lzer, Udo}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/04/Kraft%20et%20al_2013_The%20Tonalness%20Spectrum.pdf}, year = {2013}, date = {2013-01-01}, urldate = {2014-01-16}, booktitle = {Proceedings of the 16th International Conference on Digital Audio Effects}, address = {Maynooth}, abstract = {The tonalness spectrum shows the likelihood of a spectral bin be- ing part of a tonal or non-tonal component. It is a non-binary measure based on a set of established spectral features. An eas- ily extensible framework for the computation, selection, and com- bination of features is introduced. The results are evaluated and compared in two ways. First with a data set of synthetically gen- erated signals but also with real music signals in the context of a typical MIR application.}, keywords = {audio, audio analysis, features, machine learning, music, music information retrieval, tonality, tonalness}, pubstate = {published}, tppubtype = {inproceedings} } The tonalness spectrum shows the likelihood of a spectral bin be- ing part of a tonal or non-tonal component. It is a non-binary measure based on a set of established spectral features. An eas- ily extensible framework for the computation, selection, and com- bination of features is introduced. The results are evaluated and compared in two ways. First with a data set of synthetically gen- erated signals but also with real music signals in the context of a typical MIR application. |
2012 |
Lerch, Alexander An Introduction to Audio Content Analysis: Applications in Signal Processing and Music Informatics Book Wiley-IEEE Press, Hoboken, 2012, ISBN: 978-1-118-26682-3. Abstract | Links | BibTeX | Tags: analysis, audio, audio signal processing, book, feature extraction, information, listening, machine learning, machine listening, music, music analysis, music informatics, music information retrieval, retrieval, signal @book{lerch_introduction_2012, title = {An Introduction to Audio Content Analysis: Applications in Signal Processing and Music Informatics}, author = {Lerch, Alexander}, url = {http://ieeexplore.ieee.org/xpl/bkabstractplus.jsp?bkn=6266785}, isbn = {978-1-118-26682-3}, year = {2012}, date = {2012-01-01}, publisher = {Wiley-IEEE Press}, address = {Hoboken}, abstract = {With the proliferation of digital audio distribution over digital media, audio content analysis is fast becoming a requirement for designers of intelligent signal-adaptive audio processing systems. Written by a well-known expert in the field, this book provides quick access to different analysis algorithms and allows comparison between different approaches to the same task, making it useful for newcomers to audio signal processing and industry experts alike. A review of relevant fundamentals in audio signal processing, psychoacoustics, and music theory, as well as downloadable MATLAB files are also included. Please visit the companion website: www.AudioContentAnalysis.org}, keywords = {analysis, audio, audio signal processing, book, feature extraction, information, listening, machine learning, machine listening, music, music analysis, music informatics, music information retrieval, retrieval, signal}, pubstate = {published}, tppubtype = {book} } With the proliferation of digital audio distribution over digital media, audio content analysis is fast becoming a requirement for designers of intelligent signal-adaptive audio processing systems. Written by a well-known expert in the field, this book provides quick access to different analysis algorithms and allows comparison between different approaches to the same task, making it useful for newcomers to audio signal processing and industry experts alike. A review of relevant fundamentals in audio signal processing, psychoacoustics, and music theory, as well as downloadable MATLAB files are also included. Please visit the companion website: www.AudioContentAnalysis.org |
2011 |
Kirchhoff, Holger ; Lerch, Alexander Evaluation of Features for Audio-to-Audio Alignment Journal Article Journal of New Music Research, 40 (1), pp. 27–41, 2011. Abstract | Links | BibTeX | Tags: audio analysis, audio-audio alignment, dynamic time warping, features, music information retrieval, pitch chroma @article{kirchhoff_evaluation_2011, title = {Evaluation of Features for Audio-to-Audio Alignment}, author = {Kirchhoff, Holger and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/04/Kirchhoff_Lerch_2011_Evaluation%20of%20Features%20for%20Audio-to-Audio%20Alignment.pdf}, doi = {10.1080/09298215.2010.529917}, year = {2011}, date = {2011-01-01}, journal = {Journal of New Music Research}, volume = {40}, number = {1}, pages = {27--41}, abstract = {Audio-to-audio alignment is the task of synchronizing two audio sequences with similar musical content in time. We investigated a large set of audio features for this task. The features were chosen to represent four different content-dependent similarity categories: the envelope, the timbre, note-onsets and the pitch. The features were subjected to two processing stages. First, a feature subset was selected by evaluating the alignment performance of each individual feature. Second, the selected features were combined and subjected to an automatic weighting algorithm. A new method for the objective evaluation of audio-to-audio alignment systems is proposed that enables the use of arbitrary kinds of music as ground truth data. We evaluated our algorithm by this method as well as on a data set of real recordings of solo piano music. The results showed that the feature weighting algorithm could improve the alignment accuracies compared to the results of the individual features.}, keywords = {audio analysis, audio-audio alignment, dynamic time warping, features, music information retrieval, pitch chroma}, pubstate = {published}, tppubtype = {article} } Audio-to-audio alignment is the task of synchronizing two audio sequences with similar musical content in time. We investigated a large set of audio features for this task. The features were chosen to represent four different content-dependent similarity categories: the envelope, the timbre, note-onsets and the pitch. The features were subjected to two processing stages. First, a feature subset was selected by evaluating the alignment performance of each individual feature. Second, the selected features were combined and subjected to an automatic weighting algorithm. A new method for the objective evaluation of audio-to-audio alignment systems is proposed that enables the use of arbitrary kinds of music as ground truth data. We evaluated our algorithm by this method as well as on a data set of real recordings of solo piano music. The results showed that the feature weighting algorithm could improve the alignment accuracies compared to the results of the individual features. |
Lerch, Alexander Software-gestützte Merkmalsextraktion für die musikalische Aufführungsanalyse Incollection von Loesch, Heinz ; Weinzierl, Stefan (Ed.): Gemessene Interpretation - Computergestützte Aufführungsanalyse im Kreuzverhör der Disziplinen, pp. 205–212, Schott, Mainz, 2011, ISBN: 978-3-7957-0771-2. BibTeX | Tags: @incollection{lerch_software-gestutzte_2011, title = {Software-gest\"{u}tzte Merkmalsextraktion f\"{u}r die musikalische Auff\"{u}hrungsanalyse}, author = {Lerch, Alexander}, editor = {von Loesch, Heinz and Weinzierl, Stefan}, isbn = {978-3-7957-0771-2}, year = {2011}, date = {2011-01-01}, booktitle = {Gemessene Interpretation - Computergest\"{u}tzte Auff\"{u}hrungsanalyse im Kreuzverh\"{o}r der Disziplinen}, pages = {205--212}, publisher = {Schott}, address = {Mainz}, series = {Klang und Begriff}, keywords = {}, pubstate = {published}, tppubtype = {incollection} } |
Ness, Steven R; Lerch, Alexander ; Tzanetakis, George Strategies for Orca Call Retrieval to Support Collaborative Annotation of a Large Archive Inproceedings Proceedings of the International Workshop on Multimedia Signal Processing (MMSP), IEEE, Hangzhou, 2011, ISBN: 978-1-4577-1434-4. Abstract | Links | BibTeX | Tags: @inproceedings{ness_strategies_2011, title = {Strategies for Orca Call Retrieval to Support Collaborative Annotation of a Large Archive}, author = {Ness, Steven R and Lerch, Alexander and Tzanetakis, George}, url = {http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6093798}, doi = {10.1109/MMSP.2011.6093798}, isbn = {978-1-4577-1434-4}, year = {2011}, date = {2011-01-01}, booktitle = {Proceedings of the International Workshop on Multimedia Signal Processing (MMSP)}, publisher = {IEEE}, address = {Hangzhou}, abstract = {The Orchive is a large audio archive of hydrophone recordings of Killer whale (Orcinus orca) vocalizations. Researchers and users from around the world can interact with the archive using a collaborative web-based annotation, visualization and retrieval interface. In addition a mobile client has been written in order to crowdsource Orca call annotation. In this paper we describe and compare different strategies for the retrieval of discrete Orca calls. In addition, the results of the automatic analysis are integrated in the user interface facilitating annotation as well as leveraging the existing annotations for supervised learning. The best strategy achieves a mean average precision of 0.77 with the first retrieved item being relevant 95% of the time in a dataset of 185 calls belonging to 4 types.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The Orchive is a large audio archive of hydrophone recordings of Killer whale (Orcinus orca) vocalizations. Researchers and users from around the world can interact with the archive using a collaborative web-based annotation, visualization and retrieval interface. In addition a mobile client has been written in order to crowdsource Orca call annotation. In this paper we describe and compare different strategies for the retrieval of discrete Orca calls. In addition, the results of the automatic analysis are integrated in the user interface facilitating annotation as well as leveraging the existing annotations for supervised learning. The best strategy achieves a mean average precision of 0.77 with the first retrieved item being relevant 95% of the time in a dataset of 185 calls belonging to 4 types. |
2010 |
Wiesener, Constantin ; Flohrer, Tim ; Lerch, Alexander ; Weinzierl, Stefan Adaptive Noise Reduction for Real-time Applications Inproceedings Proceedings of the 128th Audio Engineering Society Convention (Preprint #8048), Audio Engineering Society, London, 2010. Abstract | Links | BibTeX | Tags: audio processing, denoising, noise reduction, real-time @inproceedings{wiesener_adaptive_2010, title = {Adaptive Noise Reduction for Real-time Applications}, author = {Wiesener, Constantin and Flohrer, Tim and Lerch, Alexander and Weinzierl, Stefan}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2015/04/Wiesener%20et%20al_2010_Adaptive%20Noise%20Reduction%20for%20Real-time%20Applications.pdf}, year = {2010}, date = {2010-01-01}, booktitle = {Proceedings of the 128th Audio Engineering Society Convention (Preprint #8048)}, publisher = {Audio Engineering Society}, address = {London}, abstract = {We present a new algorithm for real-time noise reduction of audio signals. In order to derive the noise reduction function, the proposed method adaptively estimates the instantaneous noise spectrum from an autoregressive signal model as opposed to the widely-used approach of using a constant noise spectrum fingerprint. In conjunction with the Ephraim and Malah suppression rule a significant reduction of both stationary and non-stationary noise can be obtained. The adaptive algorithm is able to work without user interaction and is capable of real-time processing. Furthermore, quality improvements are easily possible by integration of additional processing blocks such as transient preservation.}, keywords = {audio processing, denoising, noise reduction, real-time}, pubstate = {published}, tppubtype = {inproceedings} } We present a new algorithm for real-time noise reduction of audio signals. In order to derive the noise reduction function, the proposed method adaptively estimates the instantaneous noise spectrum from an autoregressive signal model as opposed to the widely-used approach of using a constant noise spectrum fingerprint. In conjunction with the Ephraim and Malah suppression rule a significant reduction of both stationary and non-stationary noise can be obtained. The adaptive algorithm is able to work without user interaction and is capable of real-time processing. Furthermore, quality improvements are easily possible by integration of additional processing blocks such as transient preservation. |
2009 |
Lerch, Alexander Software-Based Extraction of Objective Parameters from Music Performances Book GRIN Verlag, München, 2009, ISBN: 978-3-640-29496-1. Abstract | Links | BibTeX | Tags: analysis, audio, content, information, music, performance, retrieval @book{lerch_software-based_2009, title = {Software-Based Extraction of Objective Parameters from Music Performances}, author = {Lerch, Alexander}, url = {http://dx.doi.org/10.14279/depositonce-2025}, isbn = {978-3-640-29496-1}, year = {2009}, date = {2009-01-01}, publisher = {GRIN Verlag}, address = {M\"{u}nchen}, abstract = {Different music performances of the same score may significantly differ from each other. It is obvious that not only the composer’s work, the score, defines the listener’s music experience, but that the music performance itself is an integral part of this experience. Music performers use the information contained in the score, but interpret, transform or add to this information. Four parameter classes can be used to describe a performance objectively: tempo and timing, loudness, timbre and pitch. Each class contains a multitude of individual parameters that are at the performers’ disposal to generate a unique physical rendition of musical ideas. The extraction of such objective parameters is one of the difficulties in music performance research. This work presents an approach to the software-based extraction of tempo and timing, loudness and timbre parameters from audio files to provide a tool for the automatic parameter extraction from music performances. The system is applied to extract data from 21 string quartet performances and a detailed analysis of the extracted data is presented. The main contributions of this thesis are the adaptation and development of signal processing approaches to performance parameter extraction and the presentation and discussion of string quartet performances of a movement of Beethoven’s late String Quartet op. 130.}, keywords = {analysis, audio, content, information, music, performance, retrieval}, pubstate = {published}, tppubtype = {book} } Different music performances of the same score may significantly differ from each other. It is obvious that not only the composer’s work, the score, defines the listener’s music experience, but that the music performance itself is an integral part of this experience. Music performers use the information contained in the score, but interpret, transform or add to this information. Four parameter classes can be used to describe a performance objectively: tempo and timing, loudness, timbre and pitch. Each class contains a multitude of individual parameters that are at the performers’ disposal to generate a unique physical rendition of musical ideas. The extraction of such objective parameters is one of the difficulties in music performance research. This work presents an approach to the software-based extraction of tempo and timing, loudness and timbre parameters from audio files to provide a tool for the automatic parameter extraction from music performances. The system is applied to extract data from 21 string quartet performances and a detailed analysis of the extracted data is presented. The main contributions of this thesis are the adaptation and development of signal processing approaches to performance parameter extraction and the presentation and discussion of string quartet performances of a movement of Beethoven’s late String Quartet op. 130. |
2008 |
Lerch, Alexander ; Weinzierl, Stefan Digitale Audiotechnik: Grundlagen Incollection Weinzierl, Stefan (Ed.): Handbuch der Audiotechnik, pp. 785–811, Springer, Berlin, 2008, ISBN: 978-3-540-34300-4. BibTeX | Tags: @incollection{lerch_digitale_2008, title = {Digitale Audiotechnik: Grundlagen}, author = {Lerch, Alexander and Weinzierl, Stefan}, editor = {Weinzierl, Stefan}, isbn = {978-3-540-34300-4}, year = {2008}, date = {2008-01-01}, booktitle = {Handbuch der Audiotechnik}, pages = {785--811}, publisher = {Springer}, address = {Berlin}, keywords = {}, pubstate = {published}, tppubtype = {incollection} } |
Lerch, Alexander Bitratenreduktion Incollection Weinzierl, Stefan (Ed.): Handbuch der Audiotechnik, pp. 849–884, Springer, Berlin, 2008, ISBN: 978-3-540-34300-4. BibTeX | Tags: @incollection{lerch_bitratenreduktion_2008, title = {Bitratenreduktion}, author = {Lerch, Alexander}, editor = {Weinzierl, Stefan}, isbn = {978-3-540-34300-4}, year = {2008}, date = {2008-01-01}, booktitle = {Handbuch der Audiotechnik}, pages = {849--884}, publisher = {Springer}, address = {Berlin}, keywords = {}, pubstate = {published}, tppubtype = {incollection} } |
Yogev, Noam ; Lerch, Alexander A System for Automatic Audio Harmonization ( Ein System für automatische Audio-Harmonisierung) Inproceedings Proceedings of the VdT International Convention (25. Tonmeistertagung), Leipzig, 2008. Abstract | Links | BibTeX | Tags: audio, chords, harmonization, hmm, voice leading @inproceedings{yogev_system_2008, title = {A System for Automatic Audio Harmonization ( Ein System f\"{u}r automatische Audio-Harmonisierung)}, author = {Yogev, Noam and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Yogev-and-Lerch-2008-A-System-for-Automatic-Audio-Harmonization-Ein-S-1.pdf}, doi = {10.1.1.148.8391}, year = {2008}, date = {2008-01-01}, booktitle = {Proceedings of the VdT International Convention (25. Tonmeistertagung)}, address = {Leipzig}, abstract = {A rule-based system for automatic melody harmonization is presented. It models the cognitive process a human arranger undergoes when confronted with the same task, namely: segmenting the melody into phrases, tagging melody notes with harmonic functions, establishing a palette of possible chords for each note, and finding the most agreeable voicing through these chords. The system is designed to be embedded in an audio framework, which synthe- sizes a four-voiced audio output using pitch-shifting techniques. Principles of classical counterpoint as well as common voice-leading conven- tions are utilized by the system. We shall outline the various phases of computa- tion, describe the rules applied in each phase, and present perspectives regarding the stylistic flexibility suggested by the system's design.}, keywords = {audio, chords, harmonization, hmm, voice leading}, pubstate = {published}, tppubtype = {inproceedings} } A rule-based system for automatic melody harmonization is presented. It models the cognitive process a human arranger undergoes when confronted with the same task, namely: segmenting the melody into phrases, tagging melody notes with harmonic functions, establishing a palette of possible chords for each note, and finding the most agreeable voicing through these chords. The system is designed to be embedded in an audio framework, which synthe- sizes a four-voiced audio output using pitch-shifting techniques. Principles of classical counterpoint as well as common voice-leading conven- tions are utilized by the system. We shall outline the various phases of computa- tion, describe the rules applied in each phase, and present perspectives regarding the stylistic flexibility suggested by the system's design. |
2006 |
Lerch, Alexander On the Requirement of Automatic Tuning Frequency Estimation Inproceedings Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), Victoria, 2006. Abstract | Links | BibTeX | Tags: frequency, tuning @inproceedings{lerch_requirement_2006, title = {On the Requirement of Automatic Tuning Frequency Estimation}, author = {Lerch, Alexander}, url = {http://dx.doi.org/10.14279/depositonce-2037}, year = {2006}, date = {2006-01-01}, booktitle = {Proceedings of the International Society for Music Information Retrieval Conference (ISMIR)}, address = {Victoria}, abstract = {The deviation of the tuning frequency from the standard tuning frequency 440 Hz is evaluated for a database of classical music. It is discussed if and under what circumstances such a deviation may affect the robustness of pitch-based systems for musical content analysis.}, keywords = {frequency, tuning}, pubstate = {published}, tppubtype = {inproceedings} } The deviation of the tuning frequency from the standard tuning frequency 440 Hz is evaluated for a database of classical music. It is discussed if and under what circumstances such a deviation may affect the robustness of pitch-based systems for musical content analysis. |
2005 |
Lerch, Alexander ; Eisenberg, Gunnar ; Tanghe, Koen FEAPI: A Low Level Feature Extraction Plugin API Inproceedings Proceedings of 8th International Conference on Digital Audio Effects (DAFX), Madrid, 2005. Abstract | Links | BibTeX | Tags: API, audio analysis, feature extraction, music information retrieval, open source @inproceedings{lerch_feapi:_2005, title = {FEAPI: A Low Level Feature Extraction Plugin API}, author = {Lerch, Alexander and Eisenberg, Gunnar and Tanghe, Koen}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Lerch-et-al.-2005-FEAPI-A-Low-Level-Feature-Extraction-Plugin-API.pdf}, year = {2005}, date = {2005-09-01}, booktitle = {Proceedings of 8th International Conference on Digital Audio Effects (DAFX)}, address = {Madrid}, abstract = {This paper presents FEAPI, an easy-to-use platform-independent plugin application programming interface (API) for the extraction of low level features from audio in PCM format in the context of music information retrieval software. The need for and advantages of using an open and well-defined plugin interface are outlined in this paper and an overview of the API itself and its usage is given.}, keywords = {API, audio analysis, feature extraction, music information retrieval, open source}, pubstate = {published}, tppubtype = {inproceedings} } This paper presents FEAPI, an easy-to-use platform-independent plugin application programming interface (API) for the extraction of low level features from audio in PCM format in the context of music information retrieval software. The need for and advantages of using an open and well-defined plugin interface are outlined in this paper and an overview of the API itself and its usage is given. |
Lerch, Alexander ; Klich, Ingmar-Leander On the Evaluation of Automatic Onset Tracking Systems Technical Report zplane.development Berlin, 2005. Abstract | Links | BibTeX | Tags: evaluation, onset detection @techreport{lerch_evaluation_2005, title = {On the Evaluation of Automatic Onset Tracking Systems}, author = {Lerch, Alexander and Klich, Ingmar-Leander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Lerch-and-Klich-2005-On-the-Evaluation-of-Automatic-Onset-Tracking-Syst.pdf}, year = {2005}, date = {2005-01-01}, address = {Berlin}, institution = {zplane.development}, abstract = {This paper summarizes the problems, definitions and requirements that are important for the evaluation of onset tracking systems for audio signals in PCM format. Different procedures and metrics for evaluation and parametrization are presented and commented. Overall, a complete methodology for the evaluation of automatic onset detection systems is proposed.}, keywords = {evaluation, onset detection}, pubstate = {published}, tppubtype = {techreport} } This paper summarizes the problems, definitions and requirements that are important for the evaluation of onset tracking systems for audio signals in PCM format. Different procedures and metrics for evaluation and parametrization are presented and commented. Overall, a complete methodology for the evaluation of automatic onset detection systems is proposed. |
2004 |
Burred, Juan José ; Lerch, Alexander Hierarchical Automatic Audio Signal Classification Journal Article Journal of the Audio Engineering Society (JAES), 52 (7/8), pp. 724–739, 2004. Abstract | Links | BibTeX | Tags: classification, hierarchical classification, music genre, music information retrieval @article{burred_hierarchical_2004, title = {Hierarchical Automatic Audio Signal Classification}, author = {Burred, Juan Jos'{e} and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Burred-and-Lerch-2004-Hierarchical-Automatic-Audio-Signal-Classification.pdf}, year = {2004}, date = {2004-01-01}, journal = {Journal of the Audio Engineering Society (JAES)}, volume = {52}, number = {7/8}, pages = {724--739}, abstract = {The design, implementation, and evaluation of a system for automatic audio signal classification is presented. The signals are classified according to audio type, differentiating between three speech classes, 13 musical genres, and background noise. A large number of audio features are evaluated for their suitability in such a classification task, including MPEG-7 descriptors and several new features. The selection of the features is carried out systematically with regard to their robustness to noise and bandwidth changes, as well as to their ability to distinguish a given set of audio types. Direct and hierarchical approaches for the feature selection and for the classification are evaluated and compared.}, keywords = {classification, hierarchical classification, music genre, music information retrieval}, pubstate = {published}, tppubtype = {article} } The design, implementation, and evaluation of a system for automatic audio signal classification is presented. The signals are classified according to audio type, differentiating between three speech classes, 13 musical genres, and background noise. A large number of audio features are evaluated for their suitability in such a classification task, including MPEG-7 descriptors and several new features. The selection of the features is carried out systematically with regard to their robustness to noise and bandwidth changes, as well as to their ability to distinguish a given set of audio types. Direct and hierarchical approaches for the feature selection and for the classification are evaluated and compared. |
Lerch, Alexander Ein Ansatz zur automatischen Erkennung der Tonart in Musikdateien Inproceedings Proceedings of the VDT International Audio Convention (23. Tonmeistertagung), Leipzig, 2004. Abstract | Links | BibTeX | Tags: @inproceedings{lerch_ansatz_2004, title = {Ein Ansatz zur automatischen Erkennung der Tonart in Musikdateien}, author = {Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Lerch-2004-Ein-Ansatz-zur-automatischen-Erkennung-der-Tonart-.pdf}, year = {2004}, date = {2004-01-01}, booktitle = {Proceedings of the VDT International Audio Convention (23. Tonmeistertagung)}, address = {Leipzig}, abstract = {Es wird ein Verfahren zur automatischen Erkennung der Tonart von Musikdateien vorgestellt. Das Verfahren analysiert mittels einer Filterbank den Tonvorrat des Eingangssignals, der in einem Tonvektor zusammenfasst wird. Dabei sind sowohl mehrstimmige als auch ein- stimmige Eingangssignale zul\"{a}ssig. Mit Hilfe eines Nearest-Neighbour-Classifiers wird anschlie{\ss} end das wahrscheinlichste Ergebnis f\"{u}r den extrahierten Tonvektor bestimmt. Parallel zur Analyse des Tonvorrats wird die Stimmh\"{o}he des Kammertons detektiert, um eine gleichbleibende Erkennungsrate f\"{u}r Signale unterschiedlicher Stimmh\"{o}he zu gew\"{a}hrleisten.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Es wird ein Verfahren zur automatischen Erkennung der Tonart von Musikdateien vorgestellt. Das Verfahren analysiert mittels einer Filterbank den Tonvorrat des Eingangssignals, der in einem Tonvektor zusammenfasst wird. Dabei sind sowohl mehrstimmige als auch ein- stimmige Eingangssignale zulässig. Mit Hilfe eines Nearest-Neighbour-Classifiers wird anschließ end das wahrscheinlichste Ergebnis für den extrahierten Tonvektor bestimmt. Parallel zur Analyse des Tonvorrats wird die Stimmhöhe des Kammertons detektiert, um eine gleichbleibende Erkennungsrate für Signale unterschiedlicher Stimmhöhe zu gewährleisten. |
2003 |
Burred, Juan José ; Lerch, Alexander A Hierarchical Approach to Automatic Musical Genre Classification Inproceedings Proceedings of the 6th International Conference on Digital Audio Effects (DAFX), London, 2003. Abstract | Links | BibTeX | Tags: classification, hierarchical classification, music genre, music information retrieval @inproceedings{burred_hierarchical_2003, title = {A Hierarchical Approach to Automatic Musical Genre Classification}, author = {Burred, Juan Jos'{e} and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Burred-and-Lerch-2003-A-Hierarchical-Approach-to-Automatic-Musical-Genre.pdf}, doi = {10.1.1.2.6582}, year = {2003}, date = {2003-01-01}, booktitle = {Proceedings of the 6th International Conference on Digital Audio Effects (DAFX)}, address = {London}, abstract = {A system for the automatic classification of audio signals according to audio category is presented. The signals are recognized as speech, background noise and one of 13 musical genres. A large number of audio features are evaluated for their suitability in such a classification task, including well-known physical and perceptual features, audio descriptors defined in the MPEG-7 standard, as well as new features proposed in this work. These are selected with regard to their ability to distinguish between a given set of audio types and to their robustness to noise and bandwidth changes. In contrast to previous systems, the feature selection and the classification process itself are carried out in a hierarchical way. This is motivated by the numerous advantages of such a tree-like structure, which include easy expansion capabilities, flexibility in the design of genre-dependent features and the ability to reduce the probability of costly errors. The resulting application is evaluated with respect to classification accuracy and computational costs.}, keywords = {classification, hierarchical classification, music genre, music information retrieval}, pubstate = {published}, tppubtype = {inproceedings} } A system for the automatic classification of audio signals according to audio category is presented. The signals are recognized as speech, background noise and one of 13 musical genres. A large number of audio features are evaluated for their suitability in such a classification task, including well-known physical and perceptual features, audio descriptors defined in the MPEG-7 standard, as well as new features proposed in this work. These are selected with regard to their ability to distinguish between a given set of audio types and to their robustness to noise and bandwidth changes. In contrast to previous systems, the feature selection and the classification process itself are carried out in a hierarchical way. This is motivated by the numerous advantages of such a tree-like structure, which include easy expansion capabilities, flexibility in the design of genre-dependent features and the ability to reduce the probability of costly errors. The resulting application is evaluated with respect to classification accuracy and computational costs. |
2001 |
Baumgarte, Frank ; Lerch, Alexander Delayed Contribution Document 6Q/18-E: Implementation of Recommendation ITU-R BS.1387 (PEAQ) Miscellaneous 2001. @misc{baumgarte_delayed_2001, title = {Delayed Contribution Document 6Q/18-E: Implementation of Recommendation ITU-R BS.1387 (PEAQ)}, author = {Baumgarte, Frank and Lerch, Alexander}, url = {http://www.musicinformatics.gatech.edu/wp-content_nondefault/uploads/2016/10/Baumgarte-and-Lerch-2001-Delayed-Contribution-Document-6Q18-E-Implementat.pdf}, year = {2001}, date = {2001-01-01}, publisher = {ITU}, keywords = {}, pubstate = {published}, tppubtype = {misc} } |
