基于遗传算法的传声器阵列位置有源校正研究

Active microphone array position calibration based on genetic algorithm

传声器阵列的位置误差会导致高精度定位应用场景中算法性能下降,为解决这一问题,该文提出了一种二维平面传声器阵列位置参数的有源校正方法。在校正声源位置已知的情况下,利用各阵元间的到达时间差均方误差之和最小化作为优化目标设计代价函数,根据遗传算法搜索阵列真实位置的全局最优解,从而估计得到传声器阵列的准确位置信息。仿真与半消声室实验结果均表明,该文提出的方法可有效估计阵列的真实位置信息,其阵列校正误差总体小于基于到达方向与特征空间的校正方法,同时该方法具有鲁棒性强、适用性广的特点,具有较高的应用价值。

The position error of microphone array will degrade the performance of high-resolution sound source localization algorithms significantly.In order to solve this problem,we propose an active calibration method for the position parameters of the two-dimensional planar microphone array.When the locations of sound sources are assumed to be known,the sum of the mean square error of the time difference of arrival between elements is used to design the cost function,and the global optimal solution of the true position parameters of the array is solved with the genetic algorithm.Results of simulation and semi-anechoic chamber experiment show that the proposed method can accurately estimate the true position of the array.Also,the array calibration errors are generally smaller than that based on the direction of arrival method and the eigenstructure-based method.Furthermore,the results show that the proposed method has strong robustness and high application value.