The University of Tokyo / System Innovation, Graduate School of Engineering
Emotion-Driven Music Data Mining
The purpose of Research:To do the research on emotional recognition algorithms of music for realizing music emotion theme and find out a possible matching method of music and emotion. Process of Research: 1.Search and revise various existing possible emotion recognizations models, and establish our own emotion model. 2.Using open source software Marsyas to extract to original features of music samples. Artificially we mark each music fragments with appropriate emotion labels. 3.Search the papers and publishments about algorithms related to emotional recognition and the progress of the latest research. Make sure we understand the algorithms and decide the appropriate algorithm to use. 4. We tried to establish the model which links the music and emotional marks. We designed the experiment through machine learning algorithms, during this process we adjusted models and algorithms, evaluated accuracy and adequacy of the model. 5.Finally, we decided the appropriate emotion model, developed the software playing the music fragment and applied for the patent. Research Result: 1.We tried and figured one matching method of music and emotion, but its accuracy is insufficient. 2.Zhejiang University acquired copyright patent, and our group members got a practical new type patent(ZL 2015 2 0138998.1).