Method for speech emotion recognition by utilizing emotion perception spectrum characteristics

Abstract

The invention relates to a method for speech emotion recognition by utilizing emotion perception spectrum characteristics. Firstly, a pre-emphasis method is used to perform high-frequency enhancementon an input speech signal, and then, the signal is converted into frequency field by using fast Fourier transform to obtain a speech frequency signal; the speech frequency signal is divided into a plurality of sub-bands by using an emotion perception sub-band dividing method; emotion perception spectrum characteristic calculation is carried out on each sub-band, wherein the spectrum characteristics comprises an emotion entropy characteristic, an emotional spectrum harmonic inclination degree and an emotional spectrum harmonic flatness; global statistical characteristic calculation is carried out on the spectrum characteristics to obtain a global emotion perception spectrum characteristic vector; finally, the emotion perceptual spectrum characteristic vector is input to an SVM classifier toobtain the emotion category of the speech signal. According to the speech psychoacoustic model principle, the perceptual sub-band dividing method is used to accurately describe the emotion state information, and emotion recognition is carried out through the sub-band spectrum characteristics, and the recognition rate is improved by 10.4% compared with prior MFCC characteristics.

Description

technical field [0001] The invention relates to the technical field of speech emotion recognition, in particular to a speech emotion recognition method of emotion perception spectrum features. Background technique [0002] Speech is the most important way for people to communicate. Speech signals not only contain rich semantic information, but also carry rich emotional states. Analyzing the emotional features in speech and using machine learning methods to identify the emotional state of speech can be applied in many scenarios, such as: in virtual reality, by recognizing human emotions to improve the naturalness of human-computer interaction; in car driving, Improve driving safety by identifying the driver's mental state; in medicine, provide diagnostic evidence by identifying the patient's mental state; in automatic customer service, improve customer service quality by identifying customer emotions. In recent years, with the rapid development of artificial intelligence and...

AI Technical Summary

Problems solved by technology

However, emotional information is more delicate than voice content, and traditional spectral features such as MFCC and LPC are difficult to express closer emotional states, such as: sadness, fear

Images