Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (34): 217-219.

• 工程与应用 • Previous Articles     Next Articles

Research of extracting pathological voice’s characteristics based on HHT And recognition

GONG Ying-ji,HU Wei-ping   

  1. College of Physics and Electronic Engineering,Guangxi Normal University,Guilin,Guangxi 541004,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-01 Published:2007-12-01
  • Contact: GONG Ying-ji

基于HHT变换的病态嗓音特征提取及识别研究

龚英姬,胡维平   

  1. 广西师范大学 物理与电子工程学院,广西 桂林 541004
  • 通讯作者: 龚英姬

Abstract: This paper mainly proposes the validity of instantaneous energy’s and instantaneous frequency’s standard deviation parameter used as the pathological voice’s characteristics extracted based on HHT,describs the process of extracting characteristics in detail,and recognizes the new characteristic parameter and the MFCC coefficient uses the DHMM model.The result from recognition shows that,A-f standard deviation parameter which based on HHT is more resultful to use on describing pathological voice,distinguish the pathological voice from normal voice more effectively.

Key words: Hilbert-Huang Transform(HHT), pathological voice, A-f standard deviation parameter, Discrete Hidden Markov Models(DHMM), Mel Frequency Cepstrum Coefficient(MFCC)

摘要: 主要介绍基于HHT变换提取的瞬时能量(A)和瞬时频率(f)的标准差参数作为病态嗓音特征参数的有效性,详细描述了A-f新特征参数的提取过程,并利用DHMM模型对A-f标准差新特征参数,与语音识别中常用的MFCC系数进行识别。识别结果表明,由HHT变换提取的A-f标准差参数更适合于描述病态嗓音,更能有效区分病态嗓音和正常嗓音。

关键词: 希尔伯特黄变换, 病态嗓音, A-f标准差参数, 离散隐含马尔可夫模型, MEL频率倒谱系数