Wavelet Threshold Denoising Experiments of Air-coupled Ultrasonic Signal

Abstract

In air-coupled ultrasonic testing, the received ultrasonic signal has poor signal to noise ratio(SNR) due to low transmission ratio, noises introduced by transmitting and receiving system, and scattering noises during the propagation of ultrasound in the media. The use of wavelet threshold denoising technique to deal with handicaps above is faced with problems of parameter selection including wavelet base, decomposition level, and threshold function. The wavelet threshold denoising experiments are carried out using the single factor analysis method based on the basic principles of wavelet theory. Different wavelet bases from different wavelet family(Daubechies, Symlet, and Coiflet), decomposition level(4-8 levels), and threshold function(soft threshold function or improved threshold function based on hard and soft threshold) are used to wavelet threshold denoising process for the realistic noisy ultrasonic signal. The impacts of different parameters on noise filtering are analyzed from the comparison of SNR and frequency spectrum of the denoised signal. From the results, denoised signal possesses higher SNR and signal amplitude using wavelet base from coiflet family. The higher the decomposition level, the higher the SNR of the denoised signal, but growth trend gradually stabilized. The impact of threshold function on the performance of denoising is not very significant, and acceptable filtering results can be acquired from soft threshold or improved threshold function.