|2nd Workshop on Immersive Communication and Broadcast Systems (ICOB '05)
|E. Benetos, M. Kotti, C. Kotropoulos, Juan José Burred, Gunnar Eisenberg, Martin Haller, Thomas Sikora
|Comparison of Subspace Analysis-Based and Statistical Model-Based Algorithms for Musical Instrument Classification
|In this paper, three classes of algorithms for automatic classification of individual musical instrument sounds are compared. The first class of classifiers is based on Non-negative Matrix Factorization, the second class of classifiers employs automatic feature selection and Gaussian Mixture Models and the third is based on continuous Hidden Markov Models. Several perceptual features used in general sound classification as well as MPEG-7 basic spectral and spectral basis descriptors were measured for 300 sound recordings consisting of 6 different musical instrument classes (piano,violin, cello, flute, bassoon, and soprano saxophone) from the University of Iowa database. The audio files were split using 70% of the available data for training and the remaining 30% for testing. Experimental results are presented to compare the classifier performance. The results indicate that all algorithm classes offer an accuracy of over 95% that outperforms the state-of-the-art performance reported for the aforementioned experiment.
|E. Benetos, M. Kotti, C. Kotropoulos: Aristotle University of Thessaloniki