Estimation of Threshold Values at Various Scales

In partial discharge testing, signals are frequently contaminated with white noise and periodic interference. To eliminate these disturbances, we employ the db6 wavelet from the commonly used Daubechies wavelet series (specifically db6) to perform a 9-level multiresolution decomposition of the signal. The implementation typically involves using MATLAB's wavedec function for decomposition and wenergy for analyzing energy distribution across scales. Based on the energy characteristics of white noise, we estimate appropriate threshold values for each decomposition level using universal thresholding (commonly via the ddencmp and wdencmp functions) or level-dependent threshold estimation. Hard thresholding is applied using the wthresh function with 'h' parameter, which zeroes out coefficients below the threshold while preserving those above it unchanged. Finally, signal reconstruction is performed using waverec to restore the denoised signal. This analytical approach significantly enhances testing accuracy and reliability, providing more comprehensive and detailed information for experimental result analysis and evaluation. The method effectively preserves signal features while removing noise through careful threshold selection at different resolution levels.