Investigation Of Sound Source Localization Algorithm Based On Microphone Array And Its Hardware Realization

With the development of Sciences and technologies, there are increasing demands for human-computer interaction, and the explosive development in human-computer interaction has been obtained not only in the science research field but also in market applications. The communication mode of the human-computer interaction have evolved from the stage when it could only be controlled with programming language by the persons with professional quality to that stage when it can be freely controlled with voice by ordinary persons. It is the prerequisite for the human-computer interaction based on vioce-control that the computer can accuately receive the voice signal sent by the controller, which requires the computer to locate the target sound source in advance. At present, the research on sound source localization algorithm based on microphone array has become a new hot spot in the field of array signal technology. However, in China, the investigation of this algorithm has been engaged in relatively late, the development level is very limited and there are only few products now. Based on the situation above, many efforts in this study have been made as follows:Firstly, on the basis of the research on the fundamental principle of PHAT-GCC time delay estimation algorithm and further analysis of its structure feature and the scope of application, the improved PHAT-GCC time delay estimation algorithm has been proposed. The endpoint detection module and the SNR estimation module have been added to the PHAT-GCC time delay estimation algorithm. The endpoint detection module is responsible to decide the type of the current signal and avoid the procession of silent channel. The amount of calculation for the algorithm was significantly decreased and the efficiency of the hardware system was enhanced. Because the parameters of the weighting function would be modified by the SNR estimation module according to the variation of SNR and the specific weighting function was employed according to voice signal with different SNR in the sound field, the peak value of the Cross Correlation function is sharpened and the accuracy of the time delay estimation is improved.Secondly, based on the hardware platform of Red Cyclone E45 development board, the sound source localization system based on microphone array was designed. According to the problems that the linear microphone array and the plane microphone array fail to locate the sound source, the space 6-dimension microphone array topological structure with ambient orientation was proposed. Thanks to the adequate logical resource and the strong digital signal procession ability of the as-utilized FPGA XC6SLX45 of SPARTAN6 series, only 5.3ms would be cost each time for this sound-location system, which have already met the real-time requirement of the system.Thirdly, it was verified that the hardware system worked well during the whole experiment. The compareision test have been conducted under different sound fields and sources of sound. The test result shows that the success ratio of the improved PHAT-GCC time delay estimation algorithm is 20% more than that of the PHAT-GCC time delay estimation algorithm, which met the requirement of this project.