Computer Music Personal Emotion Analysis Model Based On Content

As an art, music is quite common in daily life. Since a long time ago, music has beena special way to express people’s emotion. Due to its diverse expression and exquisiteclassification, people are attracted and fond of it. Therefore, music emotion analysis hasbecome an important issue. And since the information age comes, music emotion analysisby computer is getting more and more popular.In general, music emotion analysis model is composed of three parts which are musiceigen model, emotion model and cognitive model. Music eigen model is composed offeatures abstracted from music pieces. In this paper, we abstracted some new features.Emotion model mainly describes emotion’s classification. We chose Hevner’s AdjectiveCircle and modified it slightly. For cognitive model we chose BP neural network.Compared with previous work, we made some progress mainly by creating some newfeatures in eigen model. Music emotion is influenced by many features such as melodyline, bass line rhythm, tempo, acoustic characters velocity and intensity. To detect theemotion that a musical piece represents, it is very important to extract some properfeatures from music and build a model based on them. In this research, some new features,including two main parts, are extracted. One of them uses the idea of ‘music area’. We canget a numerical sequence describing the music trend by calculating every time unit’smusic area value and work on the sequence using some statistical formulas. The other oneuses the method of separating notes into treble, middle and base section. Then we can getthree numerical sequences. We can work on them using some statistical formulas as well.For emotion model, we proposed our new way of using Hevner’s Adjective Circle. Besides, we proposed the personal cognitive model based on music’s characteristics.The test results prove good accuracy of this method. Finally, the result and the futurework are discussed.