摘要
为了解决噪声抑制时出现语音失真的问题,提高语音增强质量.提出一种差分波束形成结合自适应时频掩蔽的双麦克风阵列语音增强算法.采用一阶时域差分波束形成技术处理双麦阵列接收的信号,充分利用时域、空域信息,抑制方向性噪声,得到差分波束形成后的输出信号,并且对畸变的语音信号进行恢复.其次,差分后语音信号仍有部分噪声残留,从语音信号中进行噪声估计,估计先验信噪比,计算自适应时频掩蔽(AM),合成增强语音.AM通过信噪比将理想二值掩蔽和理想比值掩蔽按一定比值结合,避免过度抑制噪声.最后,使用SNR、PESQ及时域波形进行评测,将本文算法与单通道和多通道算法进行对比,实验结果证明,以上算法显著地提高了信噪比,语音质量优于对比算法,鲁棒性高且易于实现.
Abstract
In order to solve the problem of voice distortion during noise suppression and improve the quality of voice enhancement,a dual-microphone array speech enhancement algorithm based on differential beamforming com-bined with adaptive time-frequency masking is proposed.The first-order time-domain differential beamforming tech-nology was used to process the signals received by the dual-mic array,and the directional noise was suppressed by making full use of the time-domain and spatial-domain information to obtain the output signal after differential beam-forming,and restore the distorted speech signal.Secondly,because the speech signal still has some noise residue after differential operation,the noise and prior signal-to-noise ratio were estimated from the voice signal first.Adaptive time-frequency masking(AM)was calculated using the signal-to-noise ratio and then synthesized to enhance speech.The AM was combined with the ideal binary masking and the ideal ratio masking by a certain ratio through the signal-to-noise ratio to avoid excessive noise suppression.Finally,SNR,PESQ and time-domain waveforms were used to evaluate and compare the proposed algorithm with single-channel and multi-channel algorithms.The experi-mental results show that the proposed algorithm significantly improves the signal-to-noise ratio,the voice quality is better than the comparison algorithm,and the robustness is high and easy to implement.
基金项目
国家自然科学基金(61961009)
广西省无线宽带通信与信号处理重点实验室项目(GXKL06200107)
桂林电子科技大学研究生教育创新项目(2022YCXS042)
桂林电子科技大学研究生教育创新项目(2021YCXS028)
NETL
NSTL
万方数据