The Contents Based Music Retrieval Method Using Audio Feature Analysis against Polyphonic Music

摘要

This paper describes the way of Music Retrieval Method based on audio feature analysis techniques which is proposed with three major new algorithms to improve performance of conventional way and implements the whole system including client and server side prototype. The first one of the major algorithms is to extract the high level melody feature from polyphonic music using harmonic structure attribute. The second one is to extract the feature and suppress the noise from user humming signal. The last one is fusing the way of methods with Dynamic Time Warp (DTW), Linear Scaling (LS) and Quantized Binary Code (QBCode). This has focused on targeting of commercial services such as music portal services, fixed stand-alone devices, mobile devices, and so on.