二次加速SRP-PHAT声源定位算法

Sound Source Localization Algorithm Based on Twice Accelerated SRP-PHAT

基于可控功率响应和相位变换(Steering response power-phase transform,SRP-PHAT)的声源定位算法具有较高的鲁棒性和准确性,特别是在强噪声和回响条件下。但SRP-PHAT类算法在工程应用中有一个巨大的障碍,那就是需要较大的运算量。为了解决该问题,本文从阵列拓扑结构和搜索策略两个方向出发,对SRP-PHAT类算法加速。首先,通过垂直布置的阵列将二维空间的搜索转化为一维空间的搜索;其次,采用层次搜索策略,由粗至精对一维空间进行搜索。仿真实验表明,采用该方法可以节约大量的计算量。

In microphone arrays application, sound source localization is a vitally important problem. Sound source localization algorithm based on steering response power-phase trans- form（SRP-PHAT） has high robustness and accuracy, especially in noisy and echo environ- ment. But SRP-PHAT algorithm requires large computation, which is a big shortcoming in ap- plication. To solve the problem, a new algorithm is proposed by optimizing arrays construction and search strategy. Two-dimensional search space is firstly converted to one-dimensional one then a hierarchic search strategy is adopted in one-dimensional search space. Simulation results show that a bulk of computational expense is saved by using the new algorithm.