Speech Emotion Recognition Based on PCA and CHMM

Abstract

In order to improve the accuracy of speech emotion recognition, a speech emotion recognition model based on continuous hidden Markov model (CHMM) was proposed. This paper analyses the emotional characteristics of speech information, extracts 33-dimensional feature parameters based on temporal sequence and builds a speech emotion recognition model used to classify five emotional states: happiness, anger, sadness, fear and calm. Through multiple sets of comparative experiments, the results show that compared with the general HMM model and feature extraction, the PCA-CHMM model can improve the ability of speech emotion recognition.