Denoising ultrasonic echo signals with generalized S transform and singular value decomposition

To remove the noise in the echo signals of ultrasonic pulse-echo testing, this paper puts forward a denoising algorithm based on the generalized S transform (GST) and singular value decomposition (SVD). Firstly, the ultrasonic echo signals were subjected to the GST, yielding the time-frequency matrix of the signals. Next, the matrix was taken as the Hankel matrix, and went through the SVD. The threshold for singular values to be zeroed was determined by the ratio between singular entropy increments. After zeroing the singular values representing the noise, the resulting denoised 2D time-frequency matrix was subjected to the inverse GST, generating the denoised echo signals. Our method was applied to denoise the simulated ultrasonic echo signals with different signal-to-noise ratios (SNRs), and compared with the wavelet soft thresholding (WST) method. The comparison show that our method outperformed the WST, especially in denoising the signals with low SNR. In addition, a scanning acoustic microscopy (SAM) system was designed for the experimental verification of our method. The C-scan image with our method was much better than that without our method. Hence, our method was proved feasible and effective.