NDE applications of compressed sensing, signal decomposition and echo estimation

In this investigation, a compressed sensing (CS) sampling scheme is closely incorporated into ultrasound signal decomposition. The CS is used to exploit the sparsity of ultrasound echo signals and thereby significantly reduce the sampling rate with 20-30 times lower than the Nyquist rate. Furthermore, the time-of-arrivals (TOAs) of dominant echoes are estimated with the sparse sampling. The estimated TOAs along with a priori information of the transducers are used for model-based signal decomposition on the incomplete ultrasonic data, where Gaussian Chirplet (GC), a commonly used echo model, is adopted. Parameters of GC echoes are estimated for pattern recognition and defect characterization in the presence of noise with SNR as low as -5 dB. Through an experimental study, the decomposition results and estimated parameters confirm the robustness and effectiveness of the proposed technique. The study has a broad range of application in signal analysis including sparse representation, parameter estimation, and defect detection.