NON-CONTACT PULSE RATE MEASUREMENT USING FACIAL VIDEOS

Pulse rate (PR) is one of the vital physiological parameters which indicates the physiological state of individuals thus proving to be an important parameter to be monitored. In the last decade, more emphasis is given to non-contact based systems that are low-cost and are easy to use. Despite these advancements, most of these systems are suitable for a lab environment in offline situations. This project presents an effective system for the estimation of a pulse rate from facial videos. A dataset of 160 videos with pulse rate has been introduced. The dataset is obtained from 20 subjects performing 4 activities in 2 lighting conditions. Each activity is captured by a smartphone camera placed on a tripod. This dataset with facial videos and pulse rate is trained on different Convolutional Neural Network (CNN) models to predict the pulse rate. Their performances were compared to obtain better results. Another method called Eulerian video magnification (EVM) was also implemented with the same dataset and the results were compared with the CNN results for better accuracy. This technology possesses a high potential in advancing personal health care and in the field of telemedicine. Additional improvements to the proposed system with regards to movement and illumination can prove to be useful in many real-time applications.