Musical Instrument Sound Classification using Deep Convolutional Neural Network

Abstract

MFCC is one of the mostly used feature in automatic musical instrument sound classification. We also used some higher order spectral features such as spectral centroid, spectral flux and spectral kurtosis along with MFCC. The MFCC features resembles human auditory system thus gives better accuracy for identification. These features together comprise to make the model robust and accurate. CNN is used here for classification purpose. It detects the features and builds a network to identify different instruments.