A Regression Approach to Music Emotion Recognition
IEEE Transactions on Audio, Speech, and Language Processing
Hi-index | 0.00 |
Expectation is an important mechanism in shaping the affective experience of music. Central to this process is the concept of musical tension. The temporal evolution of tension during a piece of music is not only responsible for eliciting emotions but may form the basis for novel time-aware search queries in music information retrieval. This paper introduces a method of modelling musical tension based on automatically computed measures of musical complexity, psychoacoustics and musical structure. The approach involves examining time-continuous annotations of tension and constructing models with a number of regression algorithms. Highest performing models when evaluated with the R2 statistic reached 0.68 with Multiple Linear Regression in a 5 dimension feature space. When independently evaluated on unseen music data the system produced an R2 of 0.64.