Performance Evaluation of DCT, FFT and DWT Basis Compressive Sensing for Guided Wave Ultrasonic Testing

Abstract

Guided wave ultrasonic testing (GWUT) is among the optimum method that has been broadly utilised for
structural health monitoring (SHM) applications. With the rapid development of wireless and real-time in SHM
fields, the implementation of compressive sensing (CS) for GWUT has received attention for better signal-processing
compression techniques. Instead of the traditional compression method, an efficient CS promises a compression
method that can sample signals lower than the Nyquist sampling rate. In CS, the signal basis is an important
parameter to be investigated that will influence the compression performance. This study transforms the GWUT
input signal into three (3) types of familiar signal basis, such as discrete cosine transform (DCT), fast Fourier
transform (FFT) and discrete wavelet transform (DWT), and the compression performance using the CS method is
examined and compared. The results reveal that among these three (3) signals basis, considered the trade-off between
quality and compressed performance, the DCT is the best basis and the sparsest transform signal for GWUT with
excellent compressed ability and quality of reconstruction. 50 to 80 percent of the M/N sampling ratio was recorded
as the best sampling ratio for DCT, with SNR values ranging from 37 dB to 44 dB.