A new modeling procedure, called the joint soft-thresholding maximum likelihood (JSTML) mthod, is proposed fo rradar scattering to vover the important practical case where the broadband scattering signal is received with a low signal-to-noise ratio (SNR). The basic idea underlying the JSTML method is first using a soft-thresholding prefiltering process to enhance SNR, and then finding the estimates of the parameters in GTD-based model throgh the maximum likelihood (ML) method. The gentic algorithm is used to over-come the issue of convergence to the local minimum. To reduce the computational load, our choice for an initial estimator is the matrix pencil method combined wiht GTD-based model, whcih provides initial estimates of all parameters including a parameterizing the frequency dependence of the scattering mechanism. he simulation results shwo that tha JSTML method has better noise performance than the ML method.
展开▼
