公共场所异常声源定位中时延估计方法研究

HB(Hassab-Boucher)加权广义互相关(generalized cross correlation based on HB weighted function,GCC-HB)是常用的时延估计方法,在环境为弱高斯噪声情况下,可获得较为精确的时延估计值用于声源定位。通过分析认为,通常公共场所异常声音是一种短时信号,背景噪声主要为粉红噪声与脉冲噪声,符合分数低阶α稳定分布(fractional lower order alpha-stable,FLOA)。在此背景噪声的低信噪比环境下,GCC-HB方法的时延估计性能急剧下降。为此,提出基于反正切变换的改进GCC-HB的时延估计方法(improved GCC-HB method based on arc tangent transform,ATAN-IHB)。该方法首先对加噪信号采用反正切变换抑制噪声中尖峰脉冲的影响,然后结合每帧的信噪比对HB加权函数进行改变,并由多帧HB加权后的峰值确定出时延估计值。理论分析和计算机仿真结果表明,所提出的方法即使在低信噪比的环境下,也可以获得比较满意的时延估计值,具有一定的实用性价值。

Time delay estimation method of abnormal sound source localization in public places

Generalized cross correlation based on Hassab-Boucher(HB) weighted function(GCC-HB) is a well known time delay estimation method.Relatively precise time delay estimation can be achieved using this method in the environment of weak Gaussian noise.This paper considers that abnormal sound signals in public places usually are short-time signals;the background noise is mainly pink noise and impulsive noise,and follows the fractional lower order alpha-stable(FLOA) distribution.The performance of GCC-HB degrades sharply under background noise with low signal to noise ratio(SNR).So,to improve the performance of GCC-HB,this paper proposes a modified GCC-HB method based on arc tangent transform(ATAN-IHB).Firstly,the proposed method makes use of arc tangent transform to pre-process the abnormal sound signal with noise;and then,the HB weighted function is changed according to the SNR of each frame;at last,the time delay estimation is determined from the peak of several frame HB weighted functions.The results of theoretical analysis and computer simulation indicate that the proposed method can get satisfactory value of time delay estimation even in the environment of low SNR with the noise of FLOA,and can be applied in practice.