Efficient feature mining in music objects

Abstract

This paper proposes novel strategies for efficiently extracting repeating patterns and frequent note sequences in music objects. Based on bit stream representation, the bit index sequences are designed for representing the whole note sequence of a music object with little space requirement. Besides, the proposed algorithm counts the repeating frequency of a pattern efficiently to rapidly extracting repeating patterns in a music object. Moreover, with the assist of appearing bit sequences, another algorithm is proposed for verifying the frequent note sequences in a set of music objects efficiently. Experimental results demonstrate that the performance of the proposed approach is more efficient than the related works.