基于保留部分频域镜像分量的声源定位算法

SOUND SOURCE LOCALISATION ALGORITHM BASED ON RETAINING PARTIAL MIRROR COMPONENTS IN FREQUENCY DOMAIN

针对传统的SRP-PHAT(Steered Response Power with Phase Transform)声源定位算法容易受噪声影响而导致定位性能降低的问题,提出一种频域补零且保留部分镜像分量的改进算法。该算法首先通过傅里叶变换将接收信号变换到频域,然后在高频端补零至20倍帧长,同时保留部分镜像分量。在此基础上计算麦克风对接收信号的互功率谱密度函数,作傅里叶逆变换得到相位变换加权的广义互相关(GCC-PHAT)函数。保留的镜像分量拓宽了信号频域,使GCC-PHAT函数的峰更为尖锐,累加后得到的SRPPHAT函数的空间谱峰也就更加尖锐,从而提高定位性能。实验表明,相比于传统算法,改进算法能显著提高定位成功率。

To deal with the problem of the sound source localisation algorithm of traditional steered response power with phase transform weighting (SRP-PHAT) that its localisation performance is easily degraded due to noise influence, in this paper we propose an improved algorithm which pads the zeros in frequency domain and retains partial mirror components as well. First, the algorithm transforms the received signals to frequency domain through fast Fourier transform ( FFT), and then pads the zeros to reach 20 times of the frame length in highfrequency band while preserving part of the mirror components. On this basis, the cross power spectral density function of the microphone pair on received signals can be estimated, and the corresponding generalised cross correlation with phase transform weighting ( GCC-PHAT) function can be obtained by taking inverse fast Fourier transform (IFFT) . The retained mirror components broaden the signal spectrum so that the peak of GCC-PHAT function becomes sharper. Consequently, the spatial spectrum peak of SRP-PHAT function, which is theaccumulation of GCC-PHAT functions for all of the microphone pairs, becomes sharper, thus the localisation performance is improved.Experiments show that compared with conventional algorithms, the proposed algorithm can considerably enhance the success rates of sound source localisation.