球形麦克风阵列时频故障信号定位算法研究  

Research on location algorithm of time-frequency fault signal based on spherical microphone array

作　　者：邓昌建[1] 蒋世奇[1] 蔚泽峰[1] 明祥林 

机构地区：[1]成都信息工程大学控制工程学院,成都610225

出　　处：《电子测量与仪器学报》2017年第2期309-314,共6页Journal of Electronic Measurement and Instrumentation

基　　金：四川省教育厅自然科学基金重点项目(14ZA0172)资助

摘　　要：针对广义互相关算法在强噪声环境下,设备故障时频声音信号定位困难的问题,提出一种结合图像分析与短时傅里叶变换(STFT)的时频信号子空间延时(TFSD)估计算法。利用STFT域中的噪声与信号特征,结合STFT域中不同麦克风测量数据的相关信息,使用滤波、谱减等方法进行信号、噪声区域分离,构建信号主成分区域,通过反演构造信号子空间,从而进行端点识别,延时值估计,实现声源定位。仿真结果表明,即使在加性噪声幅度远大于故障信号的2倍,也能实现准确的延时估计,延时估计误差和5 m×5 m范围的定位误差小于10%,实测也证实了该结果的正确性。To solve the location problem of the time-frequency sound signal of equipment fault under the high noise environment when using algorithm of GCC,the time-frequency subspace delay（ TFSD） estimation algorithm which combine short-time Fourier transform（ STFT） domain and image analysis is proposed. The characteristics of noise and signal in STFT domain is used,meanwhile the correlation information of different microphone is considered,the principle signal subspace is formed using the method of filter or spectral subtraction to separate area of signal and noise subspace. Based on signal subspace,the signal in time domain is inversed,then the endpoint detection is undertaken,the delay is estimated and the sound source is located. The simulation results demonstrate that the time delay can be estimated accurately even the amplitude of additive noise is more than two times of signal,the estimation error of delay and location in area of 5 m × 5 m,this results are proved by measurement.

关 键 词：麦克风阵列 短时傅里叶变换域 信号子空间 反演 

分 类 号：TN929.52[电子电信—通信与信息系统]