摘要
目前盲分离算法相当复杂,对硬件的运算速度要求很高,无法满足实时性要求。根据盲信号分离的原理和特点,提出了一种基于延时估计的实时语音盲分离算法,大大提高了盲解分离的运算速度,主客观评价表明其具有很好的分离效果,证明了算法的有效性。
The presented blind source separation algorithms are very complicated for the hardware with the realtime operation. According to the principles and characteristics of the blind source separation, a real-time blind speech separation algorithm which can improve both the separation speed and the separation performance is proposed. The simulating results show that the proposed algorithm is effective.
出处
《电声技术》
2007年第2期47-50,共4页
Audio Engineering
基金
国家自然科学基金(60472103)
上海市优秀学科带头人基金(05XP14027)
上海市重点学科项目(T0102)
关键词
盲信号分离
延时估计
联合近似特征矩阵对角化算法
blind source separation
time-delay estimation
joint approximate diagonalizafion of eigen-matrices algorithm
