Research Of DOA Estimation Algorithm Under The Background Of Microphone Array Error

As a key technology in the field of signal processing, microphone array has been proved superior in video conference, voiceprint recognition and so on.Meanwhile MUSIC algorithm, which is one of classical super-resolution algorithms,has high accuracy because of breaking the Rayleigh limit. However, DOA estimation requires higher accuracy, faster system response, stronger adaptability in current environment with " Multipath Fading " ," Multiple Access Interference " ,so that MUSIC algorithm should be improved. However, amplitude and phase or position error have an influence on DOA estimation accuracy. Therefore, in this paper, DOA estimation algorithm is researched considering amplitude and phase or position error.Main work of this paper includes:(1). MUSIC algorithm is improved. MSCS(MUSIC Symmetric Compression Spectrum) is adapted to deal with large amount of calculation in MUSIC algorithm.When MSCS algorithm searches near the spectrum peak, search range of a small area is redefined at the extreme point, and MUSIC algorithm is used to continue the peak search in this range ,so as to effectively reduce the computational complexity and improve the accuracy of the algorithm. Simulation result shows that the accuracy of this algorithm is consistent with that of MUSIC algorithm with the higher SNR(about 25dB or more), and the system response time is reduced to 2/3. It can be seen that the performance of this algorithm is greatly improved.(2) DOA estimation with amplitude and phase error in microphone array is researched in this part. This method of microphone array self-correction does not need to obtain the exact position of the sound source just as in the conventional method. It is only necessary to place a correction sound source in the both side of the array in a mirror symmetry way, and then after some operations like matrix decomposition error can be estimated. Experiment in shows that the estimated deviation of this self-calibration method is very small even in the speech field with 0-4000Hz frequency where amplitude and phase error is not big.(3) DOA estimation with array position error in microphone array is researched in this part. In the study of array position error correction, the idea of genetic algorithm is introduced. The initial value of estimated angle is obtained by MUSIC algorithm in the case of ideal array. And the fitness of the individual in the initial group is compared to obtain the optimal individual evaluation, and then through the selection, mutation, cross and other operations to produce the next generation of search groups, repeat the evaluation process and reduce the deviation factor until the best individual is found. Finally, estimated angle is got through the MUSIC formula.Experiments and simulation results show that the maximum estimated deviation is only 0.1cm, the vast majority of the estimated deviation of the location are below 0.02cm.