计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (32): 151-154.

• 数据库、信号与信息处理 • 上一篇    下一篇

基于高斯混合模型的咳嗽音检测方法

石 锐1,王 博1,何庆华2   

  1. 1.重庆大学 计算机学院,重庆 400030
    2.第三军医大学 大坪医院野战外科研究所,重庆 400042
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-11-11 发布日期:2011-11-11

Cough sound detection algorithm based on Gaussian mixture model

SHI Rui1,WANG Bo1,HE Qinghua2   

  1. 1.College of Computer Science,Chongqing University,Chongqing 400030,China
    2.Daping Hospital Research Institute of Field Surgery,Third Military Medical University,Chongqing 400042,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-11-11 Published:2011-11-11

摘要: 快速准确地检测出采集录音中的咳嗽部分对许多呼吸道疾病的临床诊断有着重要意义。使用梅尔频率倒谱系数(MFCC)作为特征参数来分析所要处理的声音信号,并用多组训练数据分别为采集录音中的咳嗽音、说话声、笑声、清喉音等数据各建立两个高斯混合模型(GMM),将每类数据得到的两个GMM进行线性组合得到最终的表示每类数据的概率模型,进而实现对咳嗽音部分的检测。在此基础上引入了小波去噪理论,分别对每段数据去噪并进行端点检测。仿真实验结果表明所提方法能够有效提高系统的识别性能。

关键词: 咳嗽音检测, 梅尔频率倒谱系数, 高斯混合模型, 线性组合, 小波去噪

Abstract: Rapid and precise detection of cough sound from continuous recordings is meaningful for clinical diagnosis of many respiratory diseases.This paper uses Mel-frequency cepstral coefficient as the classification feature to analyze the sound signal to be processed and creates two corresponding Gaussian mixture models for the cough sound,speech voice,laughter and throat clearing sound in the recordings respectively using multiple groups of training data,then the ultimate probability models are acquired through the means of linear combination of the two GMMs of each class.Furthermore,the theory of wavelet denoising is introduced to denoise each sound signal and then detect its endpoints.Simulation experimental results indicate that the proposed method can effectively improve the performance of the detection.

Key words: cough sound detection, Mel-frequency cepstral coefficient, Gaussian mixture model, linear combination, wavelet denoise