Research And DSP Implemetation Of Sound Source Localization Based On Beam-forming Algorithm

The sound source location has a wide application in the industrial noise location, military and other fields, the beam-forming algorithm based on microphone array is the main method of sound source localization. The sound source localization based beam-forming algorithm is to obtain the position of the sound source by using microphone array collecting the sound signals and beam-forming algorithm processing sound data. Currently, the sound source localization systems based on microphone array mostly use data acquisition card, they have problems of large volume, high cost and inconvenient to use. MVDR algorithm has strong adaptability, it can work under the low SNR environment like mechanical fault detection. However, MVDR algorithm has a large amount of calculation and difficult to meet the real-time feature in general embedded system. Against the problems of large amount of data and complex calculation in the procession of beam-forming, this paper selects the DSP chip TMS320DM642 produced by TI company, it realizes the application of uniform circular array of eight microphones on TMS320DM642 based MVDR algorithm, the system has features of small size, low cost and high real-time performance.Researching source localization system of small size and high real-time performance has profound significance to industrial noise localization and military and other fields. This paper introduces the concept of noise, then makes theoretical research on sound source localization based beam-forming algorithm and describes the advantages and disadvantages of three common sound source localization algorithms. This paper selects the beam-forming algorithm as core algorithm of source localization system and compares the conventional beam-forming algorithm and MVDR algorithm by MATLAB simulation, then tests by experiments, the MVDR algorithm has higher accuracy. Against the MVDR algorithm can only locate the azimuth angle of the range of 180 degrees effectively, this paper judges position of the target sound source by using digital correlative method and transforms array sequence correspondingly, then takes advantage of MVDR algorithm for sound source localization, the azimuth angle of range of 360 degrees can be located. Comparing two times calculation, it can reduce the computation of 1/3 and improve system’s real-time performance.On the basis of research of sound source localization based on beam-forming algorithm, this paper designs uniform circular microphone array of 8 microphone element to collect sound signals and selects TMS320DM642 and AD7606 to design sound source localization system. The 8-channel sound signals can be synchronized acquisition, with a 16bits sampling precision and 25KSPS sampling rate, TMS320DM642 microprocessor with 720MHz clock frequency meet the sound signal acquisition and processing requirements absolutely.In order to improve efficiency of the software, this paper programs by using C and assembly language mixed and transfer DSPLIB library functions to calculate FFT and uses the improved MVDR algorithm. In order to more clearly observe sound source localization result, it develops display interface of the PC software. The whole system is tested by experiments repeatedly, the experiments show that the sound source localization system can locate sound source position accurately so that the deviation can be controlled within a small range and has features of miniaturization and high real-time performance.