Influence of Low-Level Features Extracted from Rhythmic and Harmonic Sections on Music Genre Classification.

We present a comprehensive evaluation of the influence of "harmonic" and rhythmic sections contained in an audio file on automatic music genre classification. The study is performed using the ISMIS database composed of music files, which are represented by vectors of acoustic parameters describing low-level music features. Non-negative Matrix Factorization serves for blind separation of instrument components. Rhythmic components are identified and separated from the rest of the audio signals. Using such separated streams, it is possible to obtain information on the influence of rhythmic and harmonic components on music genre recognition. Further, the "original" audio feature vectors stemming from the non-separated signal are extended with such that base exclusively on drum and harmonic sections. The impact of these new parameters on music genre classification is investigated comparing the "basic" k-Nearest Neighbor classifier and Support Vector Machines.