Audio retrieval using timbral feature

The increase in availability of music information demands for the development of tools for audio retrieval. Audio information retrieval implicates the retrieval of similar audio files based on the feature. Feature extraction is one of the important tasks where the entire retrieval system relies on. In this work, audio information retrieval has been performed on GTZAN datasets using Delta Mel-Frequency Cepstral Coefficients (MFCC) feature which is a kind of timbre feature. The results obtained for the various stages of feature extraction and retrieval performance plot has been presented. The average precision and recall values obtained are 78.67% and 58.02%, respectively.