Research On Robustness Speech Enhancement Method Based On Spherical Microphone Array

Speech enhancement based on microphone array combines the speech signal processingand array signal processing technologies, which realizes the suppression of interference andnoise signals to obtain the maximum extraction of the information in the expected speechsignal in the environment. Spherical microphone array has unique space symmetrycharacteristics, so it can be better used in3D space beamforming, acoustic source localization,the noise control and acoustic field space characteristic analysis etc. The research forenhancement algorithm based on spherical microphone array has become the hot spot in thefield of speech signals processing in recent years.Robustness of array is described the anti-interference ability and the fault toleranceability of the system in the complicated environment, which is used to measure theperformance of beamforming. This thesis describes the relevant basic knowledge of speechsignal processing and array signal processing, in which spherical microphone arraybeamforming method. Some theories and methods of spherical microphone array processingare summed up and analyzed. These beamforming methods include regular beamforming,delay-and-sum beamforming, Chebyshev beamforming, minimum variance distortionlessresponse beamforming and multi-constrained optimization control robustness beamforming.Considering the goal of speech enhancement beamforming is to be sure that the expectedspeech signals are located in mainlobe in the beam pattern, and interference signals and noisesignals as far as possible are located in the sidelobe even in the directions of notch, thus toachieve maximum effect of speech enhancement. With the known direction of expectedsource wave arrival, the white noise gain is adopted to guarantee the robustness of array, andthrough the linear constraint on the gains at directions of mainlobe and sidelobes, a robustnessspherical microphone array beamforming method based on linear constraint is proposed inthis thesis, in order to realize the mainlobe width and the directions of sidelobe notch can bebalanced controlled. The simulation results show that, it can not only guarantee the robustnessof array and the distortionless output in mainlobe but also achieve a better sidelobe control.