Washington -: A new computer program can measure a personís pulse by analyzing videos of them trying to hold still and spotting a tiny tic that betrays every heartbeat.
Not yet tested in a clinical setting, the algorithm could provide a way to check the health of newborns and elderly people with easily damaged skin, Fox News reported.
A camera feeding into the program could, in principle, monitor someone continuously.
Guha Balakrishnan, a Massachusetts Institute of Technology graduate student who presented his teamís project at the IEEE Computer Vision Pattern Recognition conference in Portland, Oregon, didnít set out to study the heart.
He planned to measure peopleís breathing rates by filming their heads moving up and down, in time with the expansion and contraction of their lungs. But then his videos revealed a subtle, intriguing spasm that occurred at regular intervals.
He had rediscovered a phenomenon known to medical science for more than 130 years.
Every time the heart squeezes, the body jumps up. Thatís because blood rushing up out of the heart is channeled downward by the aorta, as well as by the blood vessels it strikes in the head.
Physics dictates that the downward forces must be counterbalanced by upward forces on the blood vessels. Thus the body -- and the head -- rise like a water-propelled rocket.
The first practical device to take a pulse by monitoring this tremor dates to 1936. Invented by American physician Isaac Starr, the ballistocardiograph looked like a bed. The twitches of a sprawled out patient rocked the bed back and forth.
Balakrishnanís 21st-century twist on the idea didnít require any lying down. Each user stared at a video camera for up to 90 seconds while doing their best not to move.
Software tracked up to 1,000 points on the face and then weeded out especially slow or fast movements tied to breathing or involuntary adjustments the head makes to keep itself balanced and upright.
Nailing down the motions caused by heart contractions took a mathematical technique developed more than century ago, called principal components analysis. It finds patterns in complicated data and is often used for face recognition algorithms. In this case, a computer program tried out different combinations of the tracked points and selected the one that moved rhythmically at the steadiest pace.