A Novel Techniques for Classification of Musical Instruments

Musical instrument classification provides a framework for developing and evaluating features for any type of content-based analysis of musical signals. Signal is subjected to wavelet decomposition. A suitable wavelet is selected for decomposition. In our work for decomposition we used Wavelet Packet transform. After the wavelet decomposition, some sub band signals can be analyzed, particular band can be representing the particular characteristics of musical signal. Finally these wavelet features set were formed and then musical instrument will be classified by using suitable machine learning algorithm (classifier). In this paper, the problem of classifying of musical instruments is addressed. We propose a new musical instrument classification method based on wavelet represents both local and global information by computing wavelet coefficients at different frequency sub bands with different resolutions. Using wavelet packet transform (WPT) along with advanced machine learning techniques, accuracy of music instrument classification has been significantly improved. Keywords: Musical instrument classification, WPT, Feature Extraction Techniques, Machine learning techniques