The temporal variation in raw video which cannot be detected by human eye contains information that can be used in the medical field for diagnosis. These temporal variations can be extracted effectively using Eulerian Video Magnification(EVM), a research work done by Michael Rubinstein et.al at MIT.
This work makes use of EVM to detect pulse rate from a normal video of a body part. The raw video sequence is taken as the input and spatial decomposition is applied followed by temporal filtering of the frames in a video sequence. The resultant signal after temporal filtering is amplified to extract the hidden details in the video sequence. Videos are the input from which heartbeat and length of beats measured by amplifying the subtle motion due to the influx of blood (Newtonian reaction) at each heartbeat. Features of various body parts are tracked in the video and decomposition of their trajectories is done to generate the motion components. Further, the component which best represents the motion due to heartbeat in the temporal frequency spectrum is chosen which is analysed and the peaks of the trajectories are identified. This work aims at implementing alow-cost pulse rate detector which can detect a heartbeat from video.