This paper presents a novel method for selecting optimal wavelets for signal denoising based on the sparsity of detail components in the wavelet domain. The method defines a mean of sparsity change (μsc) parameter to quantify the separation between signal and noise coefficients in the wavelet domain. The efficacy of the method is tested on simulated and experimental signals from electron spin resonance (ESR) spectroscopy at various signal-to-noise ratios (SNRs). The results show that μsc values of signals vary abruptly between wavelets, while noise values remain similar across wavelets. For low SNR data, the change in μsc between the highest and second highest values is approximately 8-10%, while for high SNR data, it is around 5%. The mean of sparsity change increases with SNR, indicating that multiple wavelets can be used for denoising, while low SNR signals can only be efficiently denoised with a few wavelets. The method selects either a single wavelet or a collection of optimal wavelets (i.e., top five wavelets) based on the highest μsc values. The code is available on GitHub and the signalsciencelab.com website. The paper discusses the importance of wavelet families, decomposition level selection, sparsity calculation, and wavelet selection criteria. It also presents results and discussions on simulated and experimental data, demonstrating the effectiveness of the method in separating noise and signal components in the wavelet domain. The method is advantageous over traditional trial-and-error approaches and can be integrated into any wavelet-based algorithm as a preprocessing tool for wavelet selection.This paper presents a novel method for selecting optimal wavelets for signal denoising based on the sparsity of detail components in the wavelet domain. The method defines a mean of sparsity change (μsc) parameter to quantify the separation between signal and noise coefficients in the wavelet domain. The efficacy of the method is tested on simulated and experimental signals from electron spin resonance (ESR) spectroscopy at various signal-to-noise ratios (SNRs). The results show that μsc values of signals vary abruptly between wavelets, while noise values remain similar across wavelets. For low SNR data, the change in μsc between the highest and second highest values is approximately 8-10%, while for high SNR data, it is around 5%. The mean of sparsity change increases with SNR, indicating that multiple wavelets can be used for denoising, while low SNR signals can only be efficiently denoised with a few wavelets. The method selects either a single wavelet or a collection of optimal wavelets (i.e., top five wavelets) based on the highest μsc values. The code is available on GitHub and the signalsciencelab.com website. The paper discusses the importance of wavelet families, decomposition level selection, sparsity calculation, and wavelet selection criteria. It also presents results and discussions on simulated and experimental data, demonstrating the effectiveness of the method in separating noise and signal components in the wavelet domain. The method is advantageous over traditional trial-and-error approaches and can be integrated into any wavelet-based algorithm as a preprocessing tool for wavelet selection.