An indoor directional speech enhancement method based on GSMDS superdirectivity beamforming
Aiming at the problems of reverberation interference and poor directivity of the main lobe during indoor direc-tional speech enhancement,an indoor directional speech enhancement method for superdirectivity beamforming based on Gram-Schmidt mode-beam decomposition and synthesis(GSMDS)is proposed.Firstly,the weighted pre-diction error is used for dereverberation,then GSMDS superdirectivity beamforming is used for directional target speech enhancement,and then the normalized minimum mean square error automatic gain control algorithm is used for gain adjustment of the enhanced speech signal to control the loudness stability,so as to complete directional en-hancement of the target speech.Among them,the gain regulation step also compares the frequency band equaliza-tion control algorithm.The numerical simulation results show that compared with conventional beamforming(CBF)method,the gain of the proposed method is improved by 10.1 dB and the maximum average gain is up to 22.7 dB.Meanwhile,the perceptual evaluation of speech quality(PESQ)value of the speech quality in the endfire direction is improved by 0.2,compared with the original received reverb speech signal.The four-element endfire array ex-periment is carried out in a decorated room.The experimental results show that this method can obtain good direc-tional enhancement effect in a decorated room.
indoor directional speech enhancementGram-Schmidt mode-beam decomposition and synthesis(GSMDS)superdirectivityautomatic gain controlendfire array
万方数据
维普
