The Designing And Implementing Of A Sound Sourse Localization System

In an intelligent meeting room,the speaker's location is an important piece of information needed for intelligent meeting room systems.The speaker's voice enhancement,post-audio and video editing,are required to use the speaker's location information.Aiming at the problem of speaker localization in the conference room environment,this paper designs a sound source localization system based on the microphone array,which realizes the real-time localization of the speaker in the conference room.In this paper,based on the characteristics of the speech signal spectrum,a two-dimensional near-field wideband MUSIC sound source localization algorithm is used to select a number of sub-bands for spectrum division of the array signals with higher energy sub-bands.Room scene within the allowable range.Through the simulation analysis of the algorithm in the two-dimensional region of the positioning error.Microphone array sound source localization system designed in this paper chooses electret microphones as sensor to convert the speaker's voice signal into electrical signal.By MAX9814 L amplifier electret microphone output of the electrical signal is amplified,you can get 60 dB gain,and in the MAX9814 L internal automatic gain control module intervention,does not appear "clipping" phenomenon.The core of the system is the TMS320F28335 DSP chip,which realizes the real-time acquisition of the array signal through the ADC module inside the chip.Because of its high sampling frequency,the error caused by asynchronous sampling can be neglected.By F28335 XINF external expansion interface,an expansion of a SRAM memory and a FLASH memory,in the system development phase,take full advantage of the advantages of improving the Harvard architecture,by modifying the official TI.cmd file,the default memory address is divided for the system,The memory area of the program with large amount of calculation and the data of the array signal occupying a larger space is moved to the expansion memory so that the complex sound source localization algorithm can run smoothly.The positioning result is stored in SD card in the form of.txt files to improve the real-time performance of the system and reduce unnecessary time loss in the process of locating the sound source.According to the characteristics of the FAT32 file system,this paper draws on the operation mode of some downloading software on the PC,"Pre-clustered chain" approach to reduce the amount of computation during sound source localization.This article focuses on the near field two-dimensional broadband MUSIC sound source localization algorithm on the F28335 designed a complete set of software processes.A large number of experiments were conducted in a small indoor conference area to verify the performance of the location algorithm and to further analyze the division and choice of subbands.