Efficient Selection of Rhythmic Features for Musical Instrument Recognition

Abstract

Various feature schemes have been proposed through acoustic study and pattern recognition research. In this paper our main intention is to investigate the performance of different rhythmic feature schemes as well as find a good rhythmic feature combination for a robust musical instrument classifier. Lots of work has been done on speech and speaker recognition. Musical instrument recognition is an important aspect of music information retrieval system. In this paper we have discussed different rhythmic features namely beat histogram, dynamic range, spectral crest facture, mean, variance. kurtosis etc.