Algorithm Research And Hardware Design Of Source Localization System Based On Microphone Array

In recent years, source localization system based on microphone array is being widely used in the videoconference and teleconference. However, the source localization algorithms need a great number of calculations. The localization accuracy has to be sacrificed in order to realize real-time applications. Aiming at these difficulties, several algorithms based on microphone array are discussed in this paper. And some improved algorithms are put forward based on the traditional algorithms. The improved algorithms reduce the computing time and improve the localization accuracy and real-time performance.Firstly, the hardware design of the source localization system based on microphone array is been presented. Multi-channel microphone signal can be sampled at a certain sampling frequency in the same time in this design. What's more, some optimizations have been made to this design about the the power consumption, circuit area and hardware cost. The optimized circuits are suitable for embedded devices. The optimized circuit design scheme is also given in this paper.Then the beam-forming source localization algorithm is discussed in this paper. After the simulation to this algorithm, we find that this algorithm needs a great number of calculations, so this algorithm is not suitable for the real-time applications. A modified algorithm named the beam-forming source localization algorithm based on sub-searching-space divided is put forward to overcome this problem. This modified algorithm needs less computation with higher accuracy. What's more, this modified algorithm only takes 2/7 of once localization time of the traditional algorithm to finish once localization.In addition, the traditional source localization algorithm based on TDOA is discussed in this paper. This algorithm needs less computation, so it is suitable for the real-time application. However, only when the sampling frequency is very high can the traditional source localization algorithm get the accuracy localization result. Then a modified algorithm named the TDOA source localization algorithm based on low sampling frequency is presented to overcome this problem. This modified algorithm still has high accuracy when the sampling frequency is very low. And in the same sampling frequency, the mean localization error of the modified algorithm is only 1/16 of the traditional TDOA algorithm. What's more, the modified algorithm only takes about 5.3ms to finish once localization, so it is suitable for the real-time applications.The testing and simulation results in this paper have been checked very strictly. The optimization of microphone data-collecting hardware circuits based on embedded devices and the TDOA source localization algorithm based low sampling frequency put forward in this paper that can reduce the hardware requirement of the source localization system. And this will expand the applied range of the source localization system.