论文部分内容阅读
现有的多数盲源分离(BSS,Blind Source Separation)算法都是假设混合系统是时不变的,然而在实际的通信系统中混合系统常常是时变的。传统的快速不动点(FastICA)算法具有快速收敛的优点,但是不能直接用于处理混合系统时变的盲源分离问题。为了提高盲源分离算法的收敛速度和对时变混合系统的跟踪性能,改进了传统FastICA算法,将混合信号分段,在各段样本中估计峭度并采用批处理的方法进行分离。仿真实验表明,改进后的FastICA算法能在时变环境中跟踪混合系统的时变,并能有效地抗多音干扰。
Most current Blind Source Separation (BSS) algorithms assume that hybrid systems are time invariant. However, hybrid systems are often time-varying in practical communication systems. The traditional fastICA algorithm has the advantage of fast convergence, but it can not be directly used to deal with the time-varying blind source separation problem in hybrid systems. In order to improve the convergence speed of blind source separation algorithm and the tracking performance of time-varying hybrid system, the traditional FastICA algorithm is improved. The mixed signal is segmented, kurtosis is estimated in each segment and separated by batch method. Simulation results show that the improved FastICA algorithm can track the time-varying of hybrid system in time-varying environment and can effectively resist polyphonic interference.