基于遗传算法的声源三维定位方法

相位变换加权的可控响应功率（SRP-PHAT）算法是一种基于麦克风阵列的鲁棒声源定位方法,该算法在有混响和噪声的环境下仍有较高的定位精度.但该算法用网格法对整个声源空间进行搜索,逐点计算其目标函数,因而总的计算量非常大,不适用于实时定位系统.针对SRP-PHAT的特点,采用遗传算法进行搜索,使总的计算量大幅度降低.仿真结果表明在混响时间为300ms,信噪比为5dB的条件下,该算法仍可达到较高的定位精度.

A sound source 3D localization method based on genetic algorithm

The Steer Response Power-Phase Transform （SRP-PHAT） is a robust sound source localization method based on microphone array.High accuracy localization estimation can be achieved in a noisy,reverberant environment using SRP-PHAT.However the algorithm performs an inefficient grid search which need computing the objective function in each grid,it cannot be applied to a real time localization system because of its expensive computation.Aiming at the characteristic of SRP-PHAT,genetic algorithm is used to search the optimum,which greatly reduces the computational complexity.The simulation results show that the proposed method still attains comparatively high localization accuracy under the conditions of 300ms reverberation time and 5 dB signal-to-noise ratio.