Numerical implementation of matching pursuit for the analysis of complex ultrasonic signals

Matching pursuit has typically been applied to ultrasonic signal analysis for the purpose of identifying or estimating discrete echoes. In this paper, a specific numerical implementation of matching pursuit designed for ultrasonic signal decomposition is proposed, consisting of the selection of a coarse set of basis functions, the search method for finding the best matching basis function, and interpolation of the basis function parameters to achieve high resolution. In addition, the use of matching pursuit is applied to the analysis of complex ultrasonic signals by interpreting the matching basis functions as characteristic wavelets. Changes in parameters of these wavelets are related to changes in the structure. The efficiency of the numerical implementation method is evaluated, and the capability of the feature extraction method for complex ultrasonic signals is demonstrated on experimental data from an aluminum plate in the context of structural health monitoring.