A Novel Method for PPG Motion Artifact Removal Based on GASF and cGAN

This paper proposes a novel method for eliminating photoplethysmography (PPG) motion artifacts. Initially, 1-D PPG signals are transformed into 2-D images. Subsequently, employing a conditional generative adversarial network (cGAN), the original PPG signals are reconstructed from those affected by motion artifacts. The resulting images are then reverted back to 1-D time series. The proposed model exhibits a PSNR of 28.187 and SSIM of 0.959, indicating a remarkable similarity between the reconstructed PPG signals and the reference original PPG signal. Notably, the proposed approach cleans a noisy PPG signal segment in just 0.13 seconds, highlighting its potential as a viable real-time solution for removing motion artifacts from biomedical signals.