Musical phrase representation and recognition by means of neural networks and rough sets

This paper discusses various musical phrase representations that can be used to classify musical phrases with a considerable accuracy. Musical phrase analysis plays an important role in music information retrieval domain. In the paper various representations of a musical phrase are described and analyzed. Also the experiments were designed to facilitate pitch prediction within a musical phrase by means of entropy-coding of music. We used the concept of predictive data coding introduced by Shannon. Encoded music representations, stored in the database, are then used for automatic recognition of musical phrases by means of Neural Networks (NN) and rough sets (RS). A discussion on obtained results is carried out and conclusions are included.