小波分析和MFCC融合的声音信号端点检测算法  

Audio endpoints detection algorithm based on wavelet analysis and MFCC

作　　者：张芝旖 姚恩涛[1] 石玉[1] 

机构地区：[1]南京航空航天大学自动化学院,南京211106

出　　处：《电子测量技术》2016年第7期62-66,共5页Electronic Measurement Technology

摘　　要：声振法检测风机叶片的脱层损伤是一种可行也比较容易实现的方法,声信号端点检测是声振法检测中比较重要的环节。为了提高端点检测的适应性和鲁棒性,研究了小波分析及MFCC参数的提取,结合二者特点,利用小波变换改进MFCC提取过程,提出新参数DWT-MFCC,并利用SVM进行声音信号端点检测。实验结果表明,相比于传统的小波参数和MFCC参数,在相同噪声环境下,DWT-MFCC具有更高的端点检测率。Acoustic method to detect wind vane delaminating damage is a kind of feasible and relatively easy to implement ,the method of acoustic signal endpoint detection is important link in acoustic resonance method to detect .In order to improve the adaptability and robustness of endpoint detection ,wavelet transform and the MFCC parameters extraction are studied .Based on these research ,a new parameter ,DWT‐MFCC ,is extracted .And using the SVM to audio endpoint detection .The experimental results show that ,compared with the traditional wavelet and MFCC parameters ,in the same noise environment ,DWT‐MFCC has a higher rate of endpoint detection .

关 键 词：小波变换 MFCC参数 DWT-MFCC SVM 端点检测 

分 类 号：TN912.3[电子电信—通信与信息系统]