融合声门波信号频谱特征的语音情感识别

摘要

In order to improve the accuracy of emotional speech recognition,the parabolic spectral parameter(PSP) and harmonic richness factor (HRF)which are frequent domain features of the glottal waveform are analyzed,and they are applicated in speech emotion recognition.First of all,acquisition the pronunciation rate and the maximum,minimum,range and average of pitch frequency,first three formant parameters,12 order Mel frequency cepstrum coefficients (MFCC) of six different emotions speech signals(angry,fear,happy,neutral,sad,surprise) to construct a feature vector,And use principal component analysis (PCA) method to reduce the vector dimensiom Then,extract PSP and HRF of the glottal waveform,and analyze the emotional expression ability of PSP and HRF;Finally,using the stacked autoencoderclassifier aims to classify the features which are traditional and have the characteristics of the glottal signal.The results show that it can achieve a higher recognition rate to combine with thethe spectrum feature of glottal waveform.%为了提高语音情感识别的准确率,本文针对新的声门波信号频谱特征抛物线频谱参数(parabolic spectralparameter,PSP)和谐波丰富因子(harmonic richness factor,HRF)进行了研究,并将其应用到语音的情感识别中.提取6种不同情感(生气、害怕、高兴、中性、悲伤和惊奇)语音信号的发音速率和短时能量、基音频率、前3个共振峰、12阶Mel频率倒谱系数(MFCC)的最大值、最小值、变化范围和平均值等常用特征构成一个特征矢量,并利用主成分分析方法降维;提取声门波信号的频谱特征PSP和HRF,并分析了PSP和HRF的情感表达能力;采用深度学习栈式自编码算法对只有常用特征以及融合了声门波信号频谱特征后的特征进行分类.结果表明:融合声门波信号频谱特征后识别率更高.