Radar Target Recognition Using A Modified FastICA Algorithm plus GAs
Hualin Liu, Wanlin Yang · 2006
Independent component analysis (ICA) is a statistical method developed from the separation of blind signal, and now it has been successfully used in many fields. In this paper, we present an effective technique combined with a modified fastICA (M-FastICA) algorithm plus genetic algorithms (GAs) for radar high-resolution range profiles (HRRPs) feature extraction. As we all know that the most time-consuming course in fastICA is to compute the Jacobian matrix. So in this modified version, several iterations of fastICA are merged into one iteration but only needs to compute the Jacobian matrix once time. Thereby the convergence velocity of fastICA is accelerated while the performance is not degraded. To demonstrate the above feature extraction algorithm, the classification experiment on three types of radar targets are evaluated. First M-fastICA is applied to extract the independent components from the HRRPs. Then GAs is used to select the optimal basis vectors and thus build a feature subspace. The results show that the introduced method can achieve better classification performance than both PCA and ICA