Open Peer Review. Open Publishing. Open Access. Open Discussion. Open Directory. Open Recommendations. Open API. Open Source.
Audio Artist Identification by Deep Neural Network
胡振, Kun Fu, Changshui Zhang
Jan 16, 2013 (modified: Jan 16, 2013)ICLR 2013 conference submissionreaders: everyone
Abstract:Since officially began in 2005, the annual Music Information Retrieval Evaluation eXchange (MIREX) has made great contributions to the Music Information Retrieval (MIR) research. By defining some important tasks and providing a meaningful comparison system, the International Music Information Retrieval Systems Evaluation Laboratory (IMIRSEL), organizer of the MIREX, drives researchers in the MIR field to develop more advanced system to fulfill the tasks. One of the important tasks is the Audio Artist Identification task, or the AAI task. We implemented a Deep Belief Network (DBN) to identify the artist by audio signal. As a matter of copyright, IMIRSEL didn't publish there data set and we had to construct our own. In our data set we got an accuracy of 69.87% without carefully choosing parameters while the best result reported on MIREX is 69.70%. We think our method is promising and we want to discuss with others.
Enter your feedback below and we'll get back to you as soon as possible.