Speech emotion recognition method based on variational modal decomposition and extreme learning machine

Abstract

The invention discloses a speech emotion recognition method based on variational modal decomposition and an extreme learning machine, and belongs to the field of artificial intelligence and speech recognition. The speech emotion recognition method first preprocesses an emotional speech signal through a variational modal decomposition method, the emotional speech signal is decomposed into a plurality of intrinsic mode function (IMF) components and a residual component, and the components can reflect the change of a original sequence more accurately and retain the emotional characteristics of the speech signals; then, the IMF components are subjected to hilbert conversion to obtain hillbert marginal spectrum characteristics of the IMF components; and in addition, the IMF components are reaggregated to obtain the speech signal removing the residual component, and then an MEL cepstrum function is extracted for the signal. The extracted new characteristics are added into a traditional speech emotional characteristic set, and an extreme learning machine model is constructed for classification and recognition. The speech emotion recognition method has the advantage of obtaining new speechcharacteristics through the variational modal decomposition. Compared with the traditional speech emotional characteristics, the characteristics have a higher recognition rate for speech emotion recognition.

Description

technical field [0001] The invention relates to the fields of artificial intelligence and speech recognition, in particular to a speech emotion recognition method based on variational mode decomposition and extreme learning machine. Background technique [0002] Among the many communication methods, voice signals are the fastest natural method for human-to-human and human-to-machine communication. Humans can even sense the speaker's emotional state from spoken communication. Speech emotion is a method of analyzing vocal behavior and refers to indicators of various influences (such as mood, emotion, and stress), focusing on the nonverbal aspects of speech. In this context, the main challenge of speech emotion recognition (SER) is to extract some objective and measurable speech feature parameters that can reflect the speaker's emotional state. In recent years, speech emotion recognition has received extensive attention in the fields of human-computer communication, robot com...

AI Technical Summary

Problems solved by technology

The existing speech emotion features still have the problem of poor performance in classification and recognition

Images