【作者】 黎妍； 杨震；

【Author】 Li Yan,Yang Zhen (Institute of Signal Processing and Transmission Nanjing University of Posts and Telecommunications;210003)

【机构】 南京邮电大学信号处理与传输研究院；

【摘要】 本文针对一般盲源分离算法中降噪程度不稳定的问题,采用自适应的思想,设计一个时变的分离矩阵,并通过估计信号反馈和分离矩阵更新来实现盲源分离。通过大量语音分离以及语音降噪实验,证明该算法在噪声能量较大的情况下仍能取得较好的语音增强效果,是一种稳定的盲源分离算法。更多还原

【Abstract】 This paper,in view of the unstable noise reduction condition in ICA algorithm,proposes an auto-adapted blind source separation algorithm.This separation algorithm utilize the repetitious feedbacks of estimate singles and iterations of the time-variant separation matrix to realize the blind source separation.Through sufficient experiments including speech separation and enhancement,we can proved this algorithm’s performance is preferable when the noise is strong,so it is one kind of stable blind source separation algorithms.更多还原