Master of Science (MS)
Music is a complicated form of communication, where creators and culture communicate and expose their individuality. After music digitalization took place, recommendation systems and other online services have become indispensable in the field of Music Information Retrieval (MIR). To build these systems and recommend the right choice of song to the user, classification of songs is required. In this paper, we propose an approach for finding similarity between music based on mid-level attributes like pitch, midi value corresponding to pitch, interval, contour and duration and applying text based classification techniques. Our system predicts jazz, metal and ragtime for western music. The experiment to predict the genre of music is conducted based on 450 music files and maximum accuracy achieved is 95.8% across different n-grams. We have also analyzed the Indian classical Carnatic music and are classifying them based on its raga. Our system predicts Sankarabharam, Mohanam and Sindhubhairavi ragas. The experiment to predict the raga of the song is conducted based on 95 music files and the maximum accuracy achieved is 90.3% across different n-grams. Performance evaluation is done by using the accuracy score of scikit-learn.
Sridharan, Anusha, "Music Similarity Estimation" (2018). Master's Projects. 607.