Parameter estimation of scattering components using matrix pencil based support vector regression

Time delay estimation is a fundamental signal-processing problem encountered in numerous applications. The backscattered signals emanating from a source and arriving at a receiver are usually represented by a model called the undamped exponential model. This model is used for narrowband signals whose time delays are usually estimated using super-resolution algorithms. For wideband signals, however, a geometric-theory-of-diffraction (GTD) based model is usually used to incorporate the frequency dependency of individual backscattered signals. In the literature, this model is approximated to allow for the application of existing super-resolution algorithms at the expense of estimation accuracy. This paper proposes a new model-independent estimation method called Matrix Pencil Based Support Vector Regression to overcome the limitations of existing methods. Simulation results show that the proposed algorithm succeeds in accurately estimating the time delays and frequency dependency parameters of the GTD-based model.