Musical instrument recognition using audio features with integrated entropy method

Abstract

Lots of Musical content are uploaded on social media daily. It is time-consuming to search content according to our choice. Musical information retrieval is one of the evolving research fields which deals with retrieving content from audio data. Musical instrument recognition is subdomain of musical information retrieval. Previous research work had mostly focused on various western instruments belonging to distinct families, such as brass, string and woodwind are classified. The purpose of this study is to classify musical instruments using audio Features with Integrated Entropy method. Monophonic recordings of solo instrument artists are used in the experiments. Audio features have taken into account temporal, spectral, the first 13 Mel-frequency Cepstral Coefficients (MFCC) and Gammatone Frequency Cepstral Coefficients(GFCC). The proposed method generates a vector that integrates entropy with extracted features. Musical instruments are classified using generated vector. For classification, a Support Vector Machine (SVM) has been used.