Model-based Parameter Estimation Method for Terahertz Signals

Parameter estimation hasn't been investigated in terahertz non-destructive testing (NDT) field due to the lack of the prior distribution knowledge of terahertz signals. In this study, a statistical model of terahertz signal is proposed to model the terahertz echo signal. The critical issue mainly focuses on how to obtain the optimal parameter estimation from the complex terahertz echo signal. Therefore, we provide two estimators: the maximum likelihood estimation (MLE) based on expectation maximization algorithm (EM), and the lasso estimation based on the sparse representation. Simulation and experiments have been implemented to analyze the fitting performance of the proposed model. The results validate the effectiveness and applicability of the statistical model, and indicate that the Lasso estimator outperforms the EM estimator for the parameter estimation of terahertz signal, which provides a new statistical distribution model and parameter estimation method for the terahertz signal in terahertz NDT.