Query by singing/hum ming (QbSH) system for polyphonic music retrieval

This paper proposes a query-by-singing/humming (QbSH) system which retrieves the most similar music information by comparing the input data with the extracted feature information from a polyphonic music such as a MP3. The performance of music retrieval system is mainly affected by the matching engine. Feature sequences extracted from polyphonic music tracks may have many errors. Therefore, the chroma-scale representation, compensation, and asymmetric DTW (Dynamic Time Warping) are adopted in the matching engine to reduce the influence of errors and improve the performance. The performance of various distance metrics are also investigated in this paper. In our implementation, the proposed QbSH system achieves the MRR (Mean Reciprocal Rank) of 0.718 for 1000 singing/humming queries when searching from a database of 450 polyphonic music tracks.