基于分布式麦克风阵列的声源定位算法

SOUND SOURCE LOCALISATION ALGORITHM BASED ON DISTRIBUTED MICROPHONE ARRAYS

为了提高相位变换加权的可控响应功率SRP-PHAT（Steered Response Power-Phase Transform）声源定位算法的性能,提出一种基于分布式麦克风阵列的改进算法。根据分布式麦克风阵列的特点,使用麦克风对接收信号的广义互相关GCC-PHAT（Generalized Cross-Correlation with Phase Transform weighting）函数的最大值来评价接收信号的质量。在传统SRP-PHAT算法的基础上,以该最大值为权重乘以每对麦克风接收信号的GCC-PHAT函数。该算法质量较高的麦克风对接收信号赋予了较大的权重,因而能提高定位性能。仿真结果表明,在信噪比低于10 dB,混响时间大于300 ms的条件下,改进算法的定位成功率比传统算法提高了2%~4%。

To improve the performance of the steered response power-phase transform（ SRP-PHAT） which is a popular sound source localisation algorithm,an improved SRP-PHAT algorithm based on distributed microphone arrays is presented. According to the characteristic of the distributed arrays,the quality of the received signal is evaluated by the peak value of the generalised cross-correlation with phase transform weighting（ GCC-PHAT） of the received signal of microphone pair. Based on traditional SRP-PHAT algorithm,this peak value is used as the weight to multiply the GCC-PHAT function of the signals each pair of microphones received. In proposed algorithm,the received signals of microphone pair with higher quality are given bigger weights,so that are able to improve their localisation performance. Simulation results show that the improved algorithm facilitates 2% ~ 4% increase in non-anomaly rate compared with the conventional algorithm under the conditions of SNR lower than 10 dB and reverberation time greater than 300 ms.