计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (22): 203-206.

• 信号处理 • 上一篇    下一篇

ITD在语音情感识别中的研究

叶吉祥1,2,刘  亚1   

  1. 1.长沙理工大学 计算机与通信工程学院,长沙 410114
    2.中南大学 信息科学与工程学院,长沙 410083
  • 出版日期:2014-11-15 发布日期:2014-11-13

Speech emotion recognition based on Intrinsic Time-scale Decomposition

YE Jixiang1,2, LIU Ya1   

  1. 1.College of Computer and Communication Engineering, Changsha University of Science and Technology, Changsha 410114, China
    2.College of Information Science and Engineering, Central South University, Changsha 410083, China
  • Online:2014-11-15 Published:2014-11-13

摘要: 为了更好地表征语音情感状态,将固有时间尺度分解(ITD)用于语音情感特征提取。从语音信号中得到前若干阶合理旋转(PR)分量,并提取PR分量的瞬时参数特征和关联维数,以此作为新的情感特征参数,结合传统特征使用支持向量机(SVM)进行语音情感识别实验。实验结果显示,引入PR特征参数后,与传统特征的方案相比,情感识别率有了明显提高。

关键词: 固有时间尺度分解, 合理旋转分量, PR特征参数, 情感识别

Abstract: In order to express speech emotional state better, this paper takes the Intrinsic Time-scale Decomposition(ITD) into extracting speech emotion features, decomposes the emotion speech into a sum of Proper Rotation(PR) components, extracts instantaneous characteristic parameters and correlation dimension as new emotional characteristic parameters, combines with traditional features and uses Support Vector Machine(SVM) for speech emotional recognition. The results show that recognition accuracy is improved obviously through using PR features parameters.

Key words: Intrinsic Time-scale Decomposition(ITD), Proper Rotation components(PR), PR features parameters, emotion recognition