雷达信号的盲分离

Blind Separation of Radar Signals

盲信号处理已成为近年信号处理和神经网络热点领域.但主要用于语言信号处理,而应用于雷达信号处理并不多见.将盲源分离算法应用于雷达阵列接收信号处理,提出了一种新的盲源分离算法的性能评价标准—相关系数法.首先研究了雷达阵列接收模型,然后分析了雷达阵列接收信号的特性.由分析可知:当雷达目标和杂波源在空间上与雷达接收机的距离基本一致时,雷达阵列接收信号在雷达接收机上的混叠是瞬时混叠;同时雷达阵列接收信号均为超高斯信号.因此可采用盲源分离算法中的定点ICA算法来分离雷达阵列接收信号.仿真结果表明,分离出来的信号与源信号的相关系数均大于0.95,说明了定点ICA算法能雷达阵列接收信号,证明了理论分析的正确性和算法的有效性.

In recent years, blind signal separation （BSS） has already become a hot field of signal processing and neural network. But it has been mainly applied to speech signal processing, and the application of BSS to radar signal processing is rare. In this paper the BSS algorithm is applied to radar array signal processing. In this paper, firstly the model of the signals received by radar array has been studied, and then the attributes of the signals received by radar array （the combination of radar clutter and the echo of the object） have been analyzed. It shows that： （1） the signals received by radar array, including radar clutter and the echo of the object, are super-Gaussian signals; （2） When the distance between the radar object and the radar receiver approximates to the distance between the source of radar clutter and the radar receiver, the signals received by radar array will arrive at the radar receiver at the same time, that is to say the model of the signals received by radar array can be regarded as an instantaneously mixed signal model. So the signals received by radar array can be separated by BSS algorithm. Then the signals received by radar array are separated by using the fixed-point independent component analysis （ICA） algorithm. In order to evaluate the performance of the BSS algorithm, the criterion based on the matrix WA （ W is the unmixing matrix, and A is the mixing matrix） has been proposed in ref. [ 1 ] （Bell and Sejnowski, 1995, Neural Computation, 7： 1 004 - 1 034）. But it can only give qualitative analysis of the performance of the BSS algorithm. So a new criterion to evaluate the performance of the BSS algorithm has been proposed which is based on the correlation coefficient between unmixed and source signals, The blind separation of radar signals studied in this paper is to extract the echo of the object from the signals received by radar array. The result of simulation given in this paper shows that all the correlation coefficients between unmixed and source signals are more than 0.95, so the signals received by radar array can be separated by fixed - point ICA algorithm successfully. To separate the singnals received by radar array by ICA algorithm can not only improve SNR, but also reserve the information of the radar clutter, so to separate the radar array received signals by ICA algorithm is more meaningful than the traditional clutter suppression.