中文摘要：

针对跳频信号的欠定盲源分离，为了解决现有的时频域方法中算法计算量大、信号存在畸变、恢复精度不高的问题，提出了一种基于滤波的跳频信号欠定盲分离算法．该算法首先根据估计到的源信号载频设计带通滤波器，利用这些滤波器对观察信号进行滤波，得到只包含某一个源信号的观测信号分量，使原分离问题分为数个时域上稀疏的欠定盲分离问题，然后对各个分量在时域上分别应用欠定盲源分离算法估计每跳数据．通过仿真对比发现，所提的滤波法得到的跳频信号更精确，信干比比时频域方法大4dB；同时所提算法处理的数据量小，计算复杂度低．

英文摘要：

Aiming at the underdetermined blind source separation （UBSS） of frequency hopping, a UB- SS algorithm based on filtering was proposed to solve the problems of high computation and low estimated accuracy of signal caused by distortion method in existing time-frequency domain. Firstly, according to the estimated carrier frequencies, some kinds of band pass filters were designed. And the observed signals were filtered by using these band pass filters. After that, we can get the filter signals just composed of one source signal. Thus, the original UBSS problem was transferred into several UBSS problems in which the sparsity in time domain is satisfied. Finally, according to the filtered signals, the source signal was estimated by using the time domain UBSS algorithm. Simulation show that proposed algorithm can get more accurate estimated signals and have fewer processing data and lower computation complexity. The signal to interference ratio o of proposed algorithm is 4 dB bigger than that of time-frequency domain method.