Performance Evaluation of Music Recommendation System Based on Onset Detection

• Main Authors: Ho-Hsiang Shih, 施賀翔
• Other Authors: Shing-Chern D.You
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/8ffc26

碩士 === 國立臺北科技大學 === 資訊工程系 === 106 === This thesis is follow up of the thesis “Music Matching using Onset Detection with RLCS Algorithm” write by the upperclassman Ro-Wei Chao. We compare the onset detection method with the music information retrieval package librosa and madmom. Evaluate the performance of onset detection and how it affect the music similarity comparison. And we do an experiment of music listening between our system and a music similarity library musly. In order to know how human’s ear feel the music similari-ty between these two systems. In addition, the program implement is transfer from matlab to python. And we use Cython and parallel computing to improve performance of RLCS algo-rithm witch is slowly and waste a lot of time.