Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast

Radar-APLANC introduces a groundbreaking approach to radar-based heartbeat sensing, addressing the challenges of noise interference and the high costs associated with labeled training data. Traditional methods often struggle with performance degradation due to noise, while learning-based techniques require expensive ground-truth physiological signals. The proposed framework leverages both heartbeat and noise ranges to create effective positive and negative samples, enhancing noise robustness. Its Noise-Contrastive Triplet loss function relies solely on these samples and pseudo-labels generated from conventional radar methods, eliminating the need for costly data. Extensive testing on the Equipleth dataset demonstrates that Radar-APLANC achieves performance on par with leading supervised methods, underscoring its potential to revolutionize non-contact health monitoring.