Popular song emotion integration and classification method

Abstract

The invention discloses a popular song emotion integration and classification method, and relates to the field of audio information processing. The method comprises the following steps: firstly, judging a music chorus appearance mode of a piece of music, and confirming different popular music modes; then, segmenting a song into N pieces of segments via a flexible segmenting method, and predictinghappiness and an intense degree of each segment; and implementing emotion integration of the whole piece of music by selecting various classifiers in accordance with the mode of a piece of popular music as well as happiness and intense degree results of the N music segments, so that an emotion label of the whole piece of music is obtained. The method provided by the invention has the advantages that V/A emotional evolution characteristics are extracted via the flexible segmenting technique and the characteristics are separately processed, so that the targeting performance of training on the popular song emotion classifiers of different structures is enhanced; and the method, which conducts song emotion classification via the structures of the popular songs and the emotional evolution characteristics, is better in effect of reflecting emotion cognition processes and features of human beings to music in comparison with methods which simply conduct integration on the basis of statistic characteristics of the whole piece of music.

Description

technical field [0001] The invention relates to an automatic pop music emotion classification method for whole songs in the field of audio information processing. Background technique [0002] The current research object of the method for song emotion classification is mostly to process a segment in the song. A basic idea is to divide the segment into fixed-length frames, directly perform emotion classification on the frame, and then count the dominant emotional types in the song segment. as sentiment type labels for song fragments. There are also frame bags [2] However, these methods do not consider the inherent characteristics of human emotional responses when listening to songs. In fact, people's emotional perception of the whole song is affected by the appearance of emotional expressions in different positions of the song, and also by the development process of emotional expressions, which are ignored by traditional frame-bag features. People have also proposed a sche...

AI Technical Summary

Problems solved by technology

In fact, people's emotional perception of the whole song is affected by the appearance of emotional expressions in different positions of the song, and also by the development process of emotional expressions, which are ignored by the traditional frame-bag feature

People have also proposed a scheme of using the chorus as the representative section to classify the emotion of the song [3] , but does not give a method for emotional synthesis based on different paragraphs

Images