Speech emotion classifying method for emotion-based characteristic optimization

Abstract

The invention discloses a speech emotion classifying method for emotion-based characteristic optimization, which comprises the following steps: (1), acquiring speech data of basic emotion states: a happy state, an angry state, an alarmed state and a sorrowful state; (2), extracting speed emotion characteristics; (3) matching emotion pairs; (4) compacting and selecting the characteristics: (4-1) reducing dimensions of linear discriminant analysis and performing respective LDA conversion by using the respective projection vector of each emotion pair, and (4-2) selecting a characteristic selecting method which is based on fisher discriminant criterion; and (5) fusing discriminations on the basis of two types of classifiers: (5-1) recording input emotion speech data, (5-2) outputting Ci and j according to the discriminations of the two types of classifiers, (5-3) obtaining the confidence coefficients wi and j of the two types of classifiers by using a formula (3), and (5-4) making a discrimination by using a correlation decoding method, wherein the discrimination result is the emotion type corresponding to the maximum correlation value.

Description

technical field [0001] The invention relates to a speech recognition method, in particular to an automatic speech emotion recognition method based on emotion-to-feature optimization. Background technique [0002] Speech emotion recognition is to recognize the speaker's emotional state from the speech signal. General speech emotion classifiers can distinguish basic emotion categories such as joy, anger, surprise, sadness, and fear to a certain extent. To build a speech emotion classifier, you first need to determine the emotion category to be recognized, establish a corresponding emotion corpus, and then find suitable speech emotion features based on the speech data in the emotion corpus, usually based on pitch, short-term energy, formant, etc. Acoustic features constructed by parameters, and finally a speech emotion model is established using statistical methods. The quality of speech emotion features has a key impact on the performance of the classifier. [0003] The per...

AI Technical Summary

Problems solved by technology

[0005] A problem in the feature optimization of speech emotion is that general speech emotion classifiers use a single set of optimal features to classify all emotion categories, and using this set of emotion features cannot achieve the classification of any two emotion categories. optimal distinction between

For example, a set of features is selected to best distinguish the five types of speech emotions A, B, C, D, and E, but this set of features often cannot minimize the misrecognition rate between A and B in an optimal sense

Images