基于听觉特性和过零点间隔的多声源定位算法

Multiple sound source localization based on auditory character and zero-crossing intervals

针对目前声源定位技术的研究现状,研究了人耳听觉系统的生理学和心理学特性,建立了基于人耳听觉系统的多声源定位模型;通过研究信号部分频率成分间的过零点时间间隔与原信号保持一致的特性,提出了用上升过零点模拟听神经的发放行为,从而引入了根据两通道间过零点时间间隔统计特征确定两耳时延的算法,并给出了噪声对时延结果影响的判定方法;相比互相关算法,该算法的计算量较小而抗噪声能力更强,声源的增多不会明显降低定位效果;通过与常用的互相关时延算法对比,仿真实验验证了该算法在较低信噪比情况下,依然可以对多个声源取得较好的定位效果。

Aiming at the current technical status of sound source localization and separation, we have studied the physiological and psychological characteristics of human auditory system, and then established a multiple sound source localization model. Through studying the statistical character of time intervals between the zero-crossings of the signals, we find that partial frequency components of the signal and the original signal have the same characteris- tics, then, propose using the upward-going zero-crossing to simulate the fire of sound nerves, and introduce a time delay algorithm based on zero-crossing time interval between channels. Compared with cross-correlation algorithm, this algorithm has less calculation tasks, stronger anti-noise ability; and increasing the number of sound sources will not significantly reduce the accuracy of multi-source localization. Simulation experiment proves that the algorithm can have a good performance in the localization of multiple sound sources under low SNR（ signal to noise ratio）.