SVM Based Classification of Sounds from Musical Instruments using MFCC Features

Abstract

This paper aims at classifying sounds obtained from different musical instruments. The proposed methodology works by extracting the Mel Frequency Cepstral Coefficients (MFCC) of a given sound signal. The extracted features are considered to be vectors input to a Support Vector Machine (SVM). The SVM classifies the MFCC feature vector of the given sound signal by using a Minimum Distance Classifier (MDC) based classification scheme which operates by calculating the Euclidean distances of the given vector from the representative pattern vectors of the different pattern classes that the SVM has been trained with. The given sound signal is then identified as being a member of the class for which the Euclidean distance is the minimum; and is thus classified.