Music genre classification by fusion of Modified Group Delay and Melodic Features

A novel method of automatic music genre classification based on the fusion of features is proposed. The features derived from the predominant melodic contour are combined with Modified Group Delay Features (MODGDF) in the front-end. Support vector machine (SVM) classifier is used for the classification of excerpts into five different musical genres. A baseline system using Mel-Frequency Cepstral Coefficients (MFCC) is also used for objective comparison. The performance is evaluated using subset of GTZAN dataset. MODGDF based system reports an overall accuracy of 71.73%. A further improvement in performance was observed when modified group delay features were combined with melodic features (accuracy of 75.50%). The results demonstrate the potential of group delay feature and melodic feature in music genre classification task.