MusicGenreClassifier

This is the implementation of the Music Genre Classifier given in this paper

• Dataset

  The GTZAN data set used is taken from this website. This includes following genres:

  • Blues
  • Classical
  • Country
  • Disco
  • Pop
  • Jazz
  • Reggae
  • Rock
  • Metal

  Each audio clip in the data set has a length of 30 seconds, are 22050Hz Mono 16-bit MPEG files.

• Features

  Following features has been used to do this classification :

  • Mel Frequency Cepstral Coeficient
  • Spectral Centroid
  • Spectral Rolloff
  • Chroma Frequency
  • Zero crossing rate

  All of the above Features, together contribute to a feature vector with 28 different features.

• Classifier

  Following classifiers are being trained :

  • Logistic regression
  • SVM
  • Neural Network