A Deep Learning-Based Remote Plethysmography with the Application in Monitoring drivers’ Wellness

In this paper we proposed a long-short term memory (LSTM) based prediction approach, called LBPA, for vital sign assessment using facial videos. The objective is to create a remote photoplethysmography (rPPG) technology which estimates the heart rate from camera, and can be utilized in driver state monitoring. Our methodology processes the facial videos to obtain accurate heart rate estimation for each frame. We evaluated the performance of our proposed method on the RGB videos from the COHFACE database and compared to the traditional machine learning techniques. The results demonstrate that the accuracy of our proposed LBPA outperforms the handcrafted approaches. The superior performance and the low error rate achieved by the LBPA (1.98 MAE and 2.88% MAPE) indicate its potential to serve as an effective tool for remote heart rate estimation, with the potential to contribute to the development of more accurate and reliable systems in monitoring the physiological state of the drivers for safety purposes.