Automatic Baseline Correction of 1D Signals Using a Parameter-Free Deep Convolutional Autoencoder Algorithm

Baseline correction techniques are highly applicable in analytical chemistry. Consequently, there is a constant demand for universal and automated baseline correction methods. Our new procedure, based on the Convolutional Autoencoder (ConvAuto) model and combined with an automated implementation algorithm (ApplyModel procedure), meets these expectations. The key advantage of this approach is its ability to handle 1D signals of various lengths and resolutions, which is a common limitation encountered in deep neural network models. The proposed procedure is fully automatic and does not require any parameter optimization. As our experiments show, the ApplyModel procedure can also be easily combined with other baseline correction methods that utilize deep neural networks, such as the ResUNet model, which also extends its practical applicability. The usability of our new approach was tested by implementing it for both simulated and experimental signals, ranging from 200 to 4000 points in length. For complex signals characterized by multiple peaks and a nonlinear background, the ConvAuto model achieved an RMSE of 0.0263, compared to 1.7957 for the ResUNet model. In the determination of Pb(II) in a certified reference material, a recovery of 89.6% was obtained, which was 1% higher than that achieved with the ResUNet model.