Study And Hardware Implementation Of Sound Source Localization Algorithm Based On Dual Microphone Arrays

Sound source localization technology is an important part in the field of speech signal processing,and has received extensive attention from researchers.Most of the research on sound source localization of microphone arrays focuses on the fixed arrays with large apertures.However,with the development of consumer electronics in recent years,embedded devices with small microphone arrays such as wireless headsets and AR glasses have become more and more popular.For embedded devices,the characteristics of limited rate,small aperture,and unfixed array spacing lead to limited accuracy of sound source localization,which affects the actual use experience.In order to solve the above problems,this thesis builds a dual-microphone array model according to the characteristics of wireless earphone equipment,studies the problem of sound source localization,and proposes an improved time delay estimation algorithm and an array calibration algorithm based on geometric solution.The proposed improved algorithm designs and implements a dual-microphone array sound source localization hardware system.Aiming at the problem of limited sampling rate and small aperture,this thesis proposes an improved time delay estimation algorithm based on the GCC-PHAT method.The accuracy of time delay estimation under low sampling rate is improved,thereby improving the accuracy of sound source localization.Compared with the traditional TDOA algorithm,the average delay estimation error under the same conditions is reduced by 87%,the average positioning error is reduced by 56%,and the total algorithm time is only increased by 0.011 ms,which meets the real-time requirements.Aiming at the problem that the array spacing is not fixed,this thesis proposes a calibration algorithm based on geometric solution.On the basis of the time delay estimation method,two groups of specific microphones are selected for geometric solution.While the spacing value is used,the time delay estimation value with relatively large error is avoided,and the accuracy of geometric calculation is improved.Compared with the calibration algorithms in the existing literature,the average calibration error under the same conditions is reduced by 51%.A dual-microphone array sound source localization system is implemented based on Xilinx FPGA platform.A preprocessing module and a digital filter module are designed for the MEMS digital microphone,and the delay estimation and geometric solution modules are designed based on the improved algorithm proposed in this thesis.The results show that the resource utilization rate of the hardware system designed in this thesis is lower than 10%,the total power consumption of the circuit under the 48 MHz system clock is 0.241 W,and the time to complete one frame calculation is 0.29ms;In an outdoor environment without reverberation,the average error of array calibration is less than 1 cm,and the average error of sound source localization is less than 4°.