基于麦克风阵列的汽车笛语识别及笛声定位方法

Recognition and localization of car whistles using the microphone array

首先采用能频积的声音检测算法检测笛声,并根据笛声信号能量和频谱特性,设计识别算法区分笛声与非笛声信号.并使用表格搜索方法辨识笛语语义,实现智能车间笛语交互.最后,结合接收信号的到达时延差和麦克风阵列的几何结构估计出鸣笛汽车的方位.实验表明,该系统的笛语识别准确度能达到90%,角度估计误差不大于3°,较好地完成了笛声源的实时定位.

Sound data collected by the microphone array is processed to realize the recognition and localization of car whistles in real time. Firstly, an algorithm based on the product of energy and frequency is adopted to detect whistles from noise. And the characteristics of the whistle signal are used to decide whether the received signal is car whistles or not. Secondly, a table-searching method can be designed to recognize the meanings of several kinds of whistles, and then enables intelligent cars to exchange some simple information in the whistle language. Finally, the deference of time delay along with the structure of the microphone array can estimate the angle of the whistles. Computer simulations show that the whistles can be recognized with high accuracy of 90%, and the error of the angle estimate is less than 3°.