Personalized Impression-Based Music Information Retrieval Method

Along with the spread of music distribution service, there are a growing interest in music information retrieval (MIR) systems. This research proposed a music search method appropriate to more suitable impression for impression by personalizing sound model used in previous research. In this research, we use data collected from 15 subjects and select songs suitable for individuals. This method presumes the impression of music by using a personal sound model and lyric model according to the rules of previous research. The impression to be estimated is based on three impressions which the subject clearly distinguished from the prior questionnaire. 14 subjects who are 20's have evaluated the impressions of three music which the proposed system predicted. The results show that more than 75% of subjects, except one impression, are consistent with the suggested impression based on the method.