Analysis On Correlations Between Emotion Related Music Elements And Acoustical Features

Music is an expression art,which transmits emotion to people through musical elements such as melody,harmony,rhythm and mode.In recent years,with the development of Internet and information technology,the demand for automatic analysis of music content through objective acoustical features has greatly increased.How to describe people’s subjective perception of musical elements from the objective dimension has also become an urgent research topic.Taking the musical elements that significantly affect people’s emotional perception of music as the main research object,this paper explores the correlation between the subjective perception and objective acoustical features of musical elements.Firstly,this paper makes partial least squares regression between the manually labeled musical elements and the scores of the 3D emotion perception model of the corresponding music segments in the Chinese and western traditional music dataset,and obtains five musical elements that significantly affect people’s music emotion perception: tempo,pulse clarity,rhythm complexity,articulation and chord complexity.Taking these five music elements as the research object,this paper extracts four kinds of features including timbre,rhythm,pitch and loudness,with a total of 1078 dimensions.These features are extracted by frequency division.Five feature selection algorithms are used for experiments.The dimension of the feature set is reduced to less than twenty dimensions through local optimal evaluation index,and the associated acoustical feature set of the five music elements is obtained.Finally,this paper uses five regression models to predict and evaluate the performance of musical elements,compares the prediction performance differences between solo and ensemble datasets,and further analyzes and explains the objective feature set corresponding to each musical element.The following main conclusions can be drawn:(1)High frequency features are more important for the prediction of performance techniques,medium frequency features are more important for the prediction of tempo,pulse clarity,rhythm complexity and chord complexity,and the prediction effect of solo music is better than that of ensemble music.(2)The three rhythm music elements of tempo,pulse clarity and rhythm complexity are highly related to the acoustical features of rhythm and timbre.The chord complexity is mainly related to the acoustical feature of pitch.(3)Onset rate and spectral contrast are common features that play an important role in the perception of musical elements.