Evaluation of mfcc estimation techniques for music similarity

Jesper Højvang Jensen, Mads Græsbøll Christensen, Manohar Murthi, Søren Holdt Jensen

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearch

    266 Downloads (Pure)

    Abstract

    Spectral envelope parameters in the form of mel-frequencycepstral coefficients are often used for capturing timbral information of music signals in connection with genre classification applications. In this paper, we evaluate mel-frequencycepstral coefficient (MFCC) estimation techniques, namely the classical FFT and linear prediction based implementations and an implementation based on the more recent MVDR spectral estimator. The performance of these methods are evaluated in genre classification using a probabilistic classifier based on Gaussian Mixture models. MFCCs based on fixed order, signal independent linear prediction and MVDR spectral estimators did not exhibit any statistically significant improvement over MFCCs based on the simpler FFT.
    Original languageEnglish
    Title of host publicationProceedings of the 14th European Signal Processing Conference
    Publication date2006
    Publication statusPublished - 2006
    EventProceedings of the 14th European Signal Processing Conference -
    Duration: 1 Jan 2006 → …

    Conference

    ConferenceProceedings of the 14th European Signal Processing Conference
    Period01/01/2006 → …

    Cite this