Research On Ultrasonic Signal Processing Based On Wavelet Analysis

Ultrasonic testing as a most widely used non-destructive testing technology is animportant mean to ensure the safe operation of equipment. Because of ultrasonic signal’snon-stationary characteristics, domain methods of signal processing, such as pure timedomain method, pure frequency domain method, Fourier transform, the windowed Fouriertransform have their own limitations, it need to use time-frequency analysis method. Thewavelet transform is a time-frequency analysis method, it has good localization properties inboth time domain and frequency domain, and is very suitable for application to ultrasonicsignal processing. Wavelet transform can be applied to ultrasonic signal’s de-noising toimprove detection accuracy, and also can be applied to the singular point positioning ofultrasonic signal. Usually ultrasonic testing suffers the impact of structural noise, and theuseful signal is submerged in noise. That brings certain difficulties for practical use. So thispaper mainly research ultrasonic signal’s de-noising.Firstly, we analyze the principle of ultrasonic testing, discuss the physical basis ofultrasound and ultrasonic target detection model. Also, we make a brief introduction to thebasic theory of wavelet analysis, including the continuous wavelet transform, discrete wavelettransform, multi-resolution analysis and fast wavelet algorithm. Specially we analyze thewavelet function, and make a detailed study on the characteristics of the wavelet function.Then, we research the ultrasonic signal’s de-noising based on wavelet analysis. There aretwo main methods: the wavelet de-noising and the wavelet packet de-noising. For the waveletde-noising, we study four aspects. First, according to the nature of the wavelet function, wechoose the suitable wavelet function for ultrasonic signal processing. Second, we analyze thedecomposition level of the ultrasonic signal, and through the de-noising experiments, we getthe optimal decomposition level. Third, we research the different threshold selection rule,analyze their strengths and weaknesses, and select the appropriate threshold. Fourth, weanalyze soft and hard threshold method and the other three improved threshold method ofultrasonic signal’s de-noising, and select the appropriate threshold function. For the waveletpacket analysis, we analyze the difference between the wavelet packet de-noising and waveletde-noising, compare their de-noising effect.Finally, as the hard threshold function is discontinuous at the threshold, thereconstruction may produce oscillations, while the soft threshold function at the threshold isdiscontinuous, and because of the constant deviation between the estimated waveletcoefficients and the true wavelet coefficients, it will affect the reconstructed signal on the approximation of the true signal. Considering the ultrasonic signal’s characteristics andthe deficiencies of the soft threshold function and the hard threshold function, this paperproposes a new ultrasonic signal processing method based on the sinusoidal thresholdfunction. The characteristics of the threshold function is showing the slow compression, withwavelet coefficients increasing, the amount of compression decreases continuously, at lastwe stop the compressing. So we can smoothly transfer from the soft threshold function to thehard threshold function. It can overcome the shortcomings of the weakness of the twofunction, and deal with the noise component in the signal effectively. Through the de-noisingsimulation experiments, we can see the sinusoidal threshold function is very suitable forultrasonic signal’s de-noising in low noise level.