Analysis of Audio Features for Music Representation

Abstract

In this paper we have evaluated the ability and importance of audio feature sets. We have used six general audio classes which includes some speech data, instrumental data and general music of different genres. The feature sets include low-level signal properties, Mel-frequency cepstral coefficients, group delay (phase based information), Chromaprint (bit level based features), audio fingerprints. The phase and magnitude features analysis is done based on the auditory perception.