Research On Sound Source Localization System Based On Time Delay Estimation Cross-power Spectrum Phase

With its advantages of strong applicability, good concealment, easy to realize and low-cost, the sound source localization technology is becoming a research focus, and is widely used in video conferencing, industrial noise reduction, security systems, robot hearing and other fields.The purpose of this dissertation is to design and produce a sound source localization system and locate the target sound source in complex real environment. In this dissertation, the basic principles of human auditory localization were introduced, several sound localization algorithms currently used were listed, the advantages and disadvantages of these algorithms were analyzed and compared first. Then, time delay estimation methods were analyzed emphatically, including the method of generalized cross correlation, the method of least mean square adaptive filter and the method of cross-power spectrum phase. This dissertation completed microphone array modeling by simulation, produced a simulated sound source based on sound recording, examined the feasibility of various time delay estimation algorithms, analyzed the sound signal spatial selectivity of microphone array, introduced various factors on the accuracy of sound source localization. In the part of system hardware, sound signal acquisition module, hardware processing module, ADC module and array signal processing module were designed. In the part of system software, based on data acquisition card drivers, time delay estimation procedure about generalized cross-correlation and procedure about cross-power spectrum phase were written, some other sound signal processing procedures were written and debugged. Software method to improve the localization accuracy was introduced in this dissertation, too.In the final experiment of sound source localization, 10 sound source positions were selected and these locations could be well detected by this system, so the localization effect to different sound source was verified. The results of the sound source localization experiment were consistent with the simulation results.