Blind Separation of Radar Signals
Xiao Wen-Shu,Zhang Xing-Gan,Du Si-Dan
DOI: https://doi.org/10.3321/j.issn:0469-5097.2006.01.006
2006-01-01
Abstract: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.