基于改进MFCC和短时能量的咳嗽音身份识别

Cough Sound Identification Based on Improved MFCC and Short-time Energy

介绍了基于咳嗽音信号的身份识别方法。针对咳嗽音信号的短时突发特点,提出了一种适合咳嗽音的改进MFCC特征参数MFCC_N,然后将MFCC_N与短时能量(E)作为组合特征参数应用于身份识别系统中。在MATLAB 7.0平台上实现基于GMM的咳嗽音身份识别系统,分别提取咳嗽音信号的MFCC、MFCC+△MFCC和MFCC_N+E作为识别参数进行对比实验。实验结果表明,采用提出的组合特征参数MFCC_N+E进行身份识别是可行有效的,与传统参数MFCC、MFCC+△MFCC相比,采用特征参数MFCC_N+E的识别系统具有较高的识别率和较低的计算复杂度。

Introduce an identification method based on cough sounds. Concerning the short burst characteristic of cough sounds, a new fea- ture parameter MFCC_N is proposed,then put MFCC_N and short-time energy together as a new series of coefficients. In order to verify the validity of the new coefficient, an identification system based on GMM is established. Respectively extract MFCC, MFCC＋ A MFCC and MFCC_N＋E as recognition parameter to compare. The experiments were implemented in MATLAB 7.0 environment. The results showed that the system performance with new parameter was feasible and could obtain higher recognition rate and 10w computational complexity than that with MFCC and MFCC＋△MFCC.