用于窄带声源定位的时延估计算法研究

Research on the algorithm of time-delay estimation applied to narrowband source location

提出了声源为窄带信号时,用一维麦克风阵列所得数据进行时延估计的算法。将由麦克风阵列采集,并通过窄带滤波器的声源信号建模为窄带信号;通过分析其频谱特征,定义了关于信号传至相邻麦克风时间差的搜索函数,以得出所求时延。对于含有背景噪声的情况,在白噪声环境下,用所得数据向量构造Hankel矩阵,通过奇异值分解法去噪;当存在窄带噪声时,用Chebyshev I低通滤波器构造出带通滤波器,达到去噪目的。最后,给出了实验仿真结果,验证方案的可行性。

In this paper,a scheme is proposed to estimate time delay,when the source is modeled as narrowband signal,with data obtained by a 1-D microphone array.The sound signal,which is passed by a narrowband filter,is modeled as narrowband signal.The characteristic of the signal in frequency domain is analyzed,and a searching function determined by the propagating time difference between two neighbor microphones is defined,in order to find the time delay.In the case of white noise,a Hankel matrix is constructed from the available data vector,and the method of singular vector decomposition is used to denoise.In the case of narrowband noise,a passband filter is constructed from Chebyshev I lowpass filter,in order to eliminate the noise.Finally,the result of simulation is presented,to evaluate the scheme.