Raga identification based on Normalized Note Histogram features

In this paper, we propose a discriminative method to identify raga of a polyphonic music clip using Normalized Note Histogram (NNH) features. The raga performed during the rendition is mainly based on the lead voice which corresponds to the main melody. In this work, the sequence of pitch values is extracted from the polyphonic music signal using salience based approach. From the sequence of pitch values, note-sequence is derived by using tonic frequency value. Note histogram is computed from note-sequence, and it is normalized by dividing each bin with the total number of voiced frames in a music clip. In this work, the sequence of bin counts in a normalized histogram is used as a feature vector for representing the music clip. The proposed classification system is designed to identify four ragas (Ahir Bhairav, Bhairavi, Bhupali and Deshkar) in open-set approach. In this paper, raga identification is carried out using template based classification approach. The proposed classifier and normalized histogram features are validated using music database consisting of 110 clips. The performance of the proposed classifier is observed to be 82.34%.