Speech Enhancement Based On Adaptive Beamforming

The speech signal inevitably suffers from noise in a complex acoustic environment,resulting in a decrease in speech signal quality and intelligibility.Therefore,speech enhancement is required.In the original speech enhancement method,the target signal is filtered while noise is removed,which severely degrades the performance of the speech processing system.With the development of technology,microphone arrays provide efficient voice processing performance.The microphone array uses spatial information to perform sound source localization firstly,and then uses a beamforming algorithm to suppress signals in other directions.The voice signal is enhanced by receiving the voice signal in the target direction with maximum efficiency.Therefore,this article contains two aspects,the sound source localization algorithm and the beamforming algorithm.For the controllable response power phase transformation sound source localization algorithm has a large amount of computation and the lack of practicality in the parameter selection in adaptive beamforming,this paper proposes an improved controllable response power phase transformation algorithm and variable step size with the microphone array as an auxiliary device.Minimum mean square adaptive beamforming algorithm.The main content of this article is as follows:Firstly,the related theory of sound source localization and beamforming algorithm is described.The classification of sound source localization is described,and some classical beamforming algorithms are briefly introduced.Secondly,an improved sound source positioning algorithm with controllable response power and phase transformation is proposed.First,remove the autocorrelation power spectrum that does not contribute to positioning,reduce the computational complexity of the algorithm,and then adopt a new coarse-to-fine search strategy to optimize the performance of the algorithm.Finally,the accuracy and minimum accuracy of the algorithm are compared before and after improvement.The root-mean-square errors and other performances verify that the improved algorithm has better performance.Again,a variable step length least mean square adaptive beamforming algorithm is proposed.For the least mean square algorithm,the problem of selecting fixed parameters is lack of practicality in practical applications.According to the relationship between the parameters and the error satisfying the function,a variable parameter method was proposed to solve the problem of the fixed parameter,and the convergence speed of the algorithm was accelerated,and the performance was better in the steady-state error.Finally,experiments show that the improved sound source localization algorithm has better robustness and better positioning performance.In terms of beamforming,the improved variable step LMS algorithm has a faster convergence rate.In adaptive beamforming,subjective and objective evaluation of the quality of the speech signal respectively validates the improved speech enhancement effect of the algorithm more obviously.