Fingertip Pulse Analysis of an Anxiety Patient Based on Deep Learning

Anxiety disorder is a common psychological condition that significantly affects individuals' well-being, including their sleep quality. This paper utilizes photoplethysmography (PPG) sensors to collect fingertip pulse data from an individual with anxiety disorder over the course of a year. Deep learning methods are then employed to explore the intrinsic correlation between the individual's anxiety levels, sleep quality, and the signals derived from the fingertip pulse. A total of 823 five-minute segments of fingertip pulse data were collected, preprocessed with denoising techniques, and segmented into pulse waveform segments based on either a 4-second or heartbeat cycle. A combination of CNN and RNN models was designed to classify the states, achieving an accuracy of 68.6% for anxiety state classification and 70.8% for sleep state classification. The results demonstrate that the use of deep learning methods enables effective discrimination of various human body states based on the fingertip pulse.