Overcoming noise in experimental signals has been an age-old problem in spectroscopy. Apart from improving instrument sensitivity, many data processing approaches such as model fitting and signal filtering methods have been applied with limited success. One of their key limitations has been recovering signals at low Signal-to-Noise Ratios (SNRs) with high fidelity. They also require a priori information and may be limited to a particular physical method. We present a new signal denoising approach based on wavelet transforms that can effectively remove noise from a signal at SNR ≧3. It incorporates the fact that wavelet transforms can be used to remove noise without a priori information about the experimental signal. Although wavelet denoising has been around for 25 years and has previously been used to remove noise from signals, it has not achieved its true potential in the past. Despite the fact that noise and signal can be distinguished and separated in the wavelet domain, standard wavelet denoising methods are unable to determine objective parameters to effectively achieve that. We developed an objective method that is able to successfully identify and separate noise from signal in the wavelet domain. This has led to the recovery of 1D signals initially at or above a SNR of 3 with excellent fidelity. The denoising method achieves more than two orders-of-magnitude improvement in (SNR) or can be used to reduce signal acquisition times by equivalent amounts. We will demonstrate the denoising procedure and its application to various types of Electron Spin Resonance signals, including representation and analysis of signal in the wavelet domain and identification and elimination of noise coefficients from signal coefficients. The method can be applied for many cases relevant in chemistry.

Data processing approach to overcome insufficient signal strength in spectroscopy

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.