PUPILHEART: HEART RATE VARIABILITY MONITORING VIA PUPILLARY FLUCTUATIONS ON MOBILE DEVICES

Abstract

Heart disease has escalated into a major public health concern despite the fact that it is preventable with early detection and treatment. Therefore, it is crucial to examine heart health every day. Seismocardiography (SCG) and photo plethysmography (PPG) are the mainstays of current mobile cardiac monitoring devices. People can't keep tabs on their hearts whenever and wherever they want to since these techniques are cumbersome and need extra equipment. We propose a method to monitor the user's heart rate by analyzing their pupillary reaction when they unlock their phone using face recognition. This method is based on our discovery of the association between pupil size and heart rate varaiability (HRV). We provide PupilHeart, a server-side and mobile terminal-based computer vision-based mobile HRV monitoring platform, with this goal in mind. At the time of phone unlock using the front-facing camera, PupilHeart records data on the user's pupil size change on the mobile terminal. The server side then pre-processes the raw pupil size data. In order to detect HRV-related time series characteristics, PupilHeart employs a 1D-convolutional neural network. The PupilHeart app also models the pupil and HRV using a three-layered recurrent neural network (RNN). With this model, PupilHeart can determine the user's heart state every time they unlock their phone by inferring their HRV. In order to thoroughly assess the efficacy of PupilHeart, we recruit 60 individuals and do both laboratory and field investigations using the prototype. All things considered, PupilHeart successfully predicts the user's HRV.