Revisiting PIC Method of Remote Respiratory Rate Measurement for Enhanced Robustness

Telehealth has revolutionized the provision of healthcare services over the past decade. In this context, various remote respiratory rate measurement methods were developed, including pixel intensity changes tracking. However, limited advancements in the intermediate signal processing stages of a widely adopted conventional approach have hindered its robust performance in real-life applications. This work proposes an innovative signal processing pipeline to the common method by introducing a new criterion for best motion signal selection based on normalized autocorrelation. By properly utilizing this and previous criteria, signals SNR was increased by 10%. Additionally, applied problems of signals grouping were investigated and solved. Evaluation results on COHFACE dataset showed 84.2% Success Rate 1, which is by 3.8% better than previous state-of-the-art method.