希尔伯特边际谱在语音情感识别中的应用

Application of Hilbert marginal spectrum in speech emotion recognition

摘要: 利用希尔伯特-黄变换(Hilbert-Huang Transform,HHT)对情感语音进行处理,得到其边际谱,然后对比分析四种情感即高兴、生气、厌恶、无情感语音信号边际谱的特征,提出四个特征量:子带能量(SE)、子带能量的一阶差分(DSE)、子带能量倒谱系数(SECC)、子带能量倒谱系数的一阶差分(DSECC)用于情感识别。用它们作说话人无关,文本无关的语音情感识别,得到最高90%的识别率,比基于傅立叶变换的梅尔频率倒谱系数(MFCC)高22个百分点。实验结果表明,基于HHT边际谱的特征能够较好地反映语音信号中的情感信息。

Abstract: Marginal spectrum of the emotional speech is obtained through Hilbert-Huang Transform.Speech signals of four different emotions,namely happy,angry,boring and natrual,are analyzed contrastively focusing on the characteristics of the marginal spectrum.Then four features:SE,DSE,SECC and DSECC are extracted for emotion recognition.Finally speaker-independent and text-independent emotion recognitions are simulated by using these features respectively,which gains the best recognition rate of 90%,which is 22 percentage higher than Fourier Transform based feature MFCC.Thus,conclusion is drawn that HHT marginal spectrum can well reflect the emotional information in speech.