A WOA-ICEEMDAN joint wavelet threshold function based denoising method for ultrasound signals
Jiaqi Zhang, Songsong Li, Xuan Liu, Haoyuan Chen, Qiaozhen Zhou · Nondestructive Testing And Evaluation · 2025
Various interference waves are often interspersed in faulty ultrasonic echo signals and are easily eradicated by noise, resulting in a low signal-to-noise ratio of the received signals. To address this problem, this study proposes a joint wavelet threshold function denoising method based on the Whale Optimization Algorithm (WOA) optimised Adaptive Complete Ensemble Empirical Decomposition of Noise (ICEEMDAN). First, the method uses WOA to optimise two parameters in ICEEMDAN: white noise amplitude weight (Nstd) and number of additions (NR). The Sample Entropy (SampEn) is combined as the fitness function, and then WOA-ICEEMDAN decomposition of the ultrasound signal is performed to obtain a series of intrinsic mode functions (IMFs). Second, the correlation coefficient method is applied to separate the IMF into useful and noise components. The multi-scale and time-frequency localisation properties of the wavelet threshold function algorithm are then utilised to analyse the noise component, extract valuable information from it, and reconstruct the processed noise and useful components to create the final denoised signal. Finally, the method is verified by simulation and real experiments. Compared with the hard and soft threshold methods, the signal-to-noise ratio improves by 44.8% and 24.9%, while the root-mean-square error declines by 52.3% and 38%, respectively.