WPD-Based Noise Reduction for Microseismic Data Through Adaptive Coefficient Shrinkage and Multi-Basis Fusion

Abstract

In microseismic monitoring, sensors distributed at different spatial positions face varying noise conditions, leading to data quality deterioration. Due to the precise time-frequency analysis capabilities, flexibility, and low computational complexity, Wavelet Packet Decomposition (WPD) has become one of the most widely used noise reduction approaches in microseismic data enhancement. However, there are still notable limitations: 1) Uniform thresholding, which uses a constant threshold across the whole time-frequency node coefficients, confuses microseismic waveform edges and noise, resulting in information loss and noise retention 2) Sensors encounter complex and varied noise characteristics from different environments, making a single wavelet basis inadequate for representing this diversity, which ultimately limits denoising performance. To overcome the above barriers, we present a denoising method based on adaptive coefficient shrinkage and dynamic weighted fusion in this work. Firstly, this method utilizes energy filtering to eliminate noise that is not temporally correlated with microseismic waveforms, then employs time-varying thresholds to refine the edges of these waveforms, effectively leveraging time and amplitude information to preserve the details of waveform edge. Secondly, a short-time feature analysis is conducted based on denoising results from various wavelet bases, implementing time-varying weights, and enhancing noise suppression while retaining waveform information. The experimental results demonstrate that the proposed method significantly improves the clarity of microseismic waveforms, offering a more effective solution for precise analysis and processing than traditional methods.

Keywords

• Adaptive coefficient shrinkage
• clustering
• dynamic weighted fusion
• microseismic
• wavelet packet decomposition