自适应降噪系统的语音分离研究
Investigation of speech separation of cross talk resistant adaptive noise canceller
-
摘要: 对抑制干扰语音的自适应降噪系统(CTRANC)进行了语音分离算法的研究。利用CTRANC抑制干扰信号的特性及语音信号的短时稳定性,借助最优控制相关理论,得到了新的语音分离方法及其自适应滤波迭代步长的计算公式。实验结果表明在双话者的情况下,这种语音分离算法能够保证自适应语音分离系统具有良好的稳定性,较好的实时跟踪能力和收敛性能,且分离出来的语音具有令人满意的清晰度。Abstract: CTRANC (cross talk resistant adaptive noise canceller) has been studied. Based on the CTRANC characteristic of suppressing the interfering signal and stability in short-term of speech signal, an adaptive speech separation algorithm based on the CTRANC system comes out. Evaluation results of objective quality measurement and informal listening test show that the new algorithm make the CTRANC speech separation system have better stability and convergence performance.