The pulse in your face and an app could replace biometric health monitors
A biometrics idea for contactless monitoring of blood pressure and other health indicators has shown up in a pair of unrelated telemedicine concepts.
Both remote diagnostic methods, one still in the labs and the other now commercialized, use AI algorithms to monitor a person’s pulse using a phone.
The techniques note changes in skin created by blood flow. As blood pulses through skin, the skin becomes more or less translucent, a biometric that can be captured in real time.
In fact, the effect can be seen by AI in any digital video.
This same effect has been proposed as a way of spotting deepfakes. At least at this point in AI’s development, realistically replicating the wash of blood flow in skin is very difficult.
University and corporate researchers say their results, which they named MetaPhys, only needs 18 seconds of video to find the subject’s pulse and heart rate. The team claims MetaPhys cuts errors 42 percent to 44 percent compared with other systems.
Home blood pressure monitoring devices are big business, and taking the cuff out of the equation could give MetaPhys a chance to compete for share in the market.
Researchers on this project included four scientists — Xin Liu, Ziheng Jiang, Xuhai Xu and Shwetak N. Patel, from the University of Washington; Josh Fromm, head of ML systems at machine learning platform vendor OctML; and Daniel McDuff, principal researcher with Microsoft Research AI.
An IEEE Spectrum article about this work points out that there are at least three ongoing similar efforts to use video for internal biometrics. One is unfolding at Oxford University, a second at Rice University and a third at Google. The Google technology is promised for Google Fit.
Meanwhile, a company called NuraLogix in Canada, is already building its Transdermal Optical Imaging into a phone product with its partner, Australian software vendor MyFiziq.
NuraLogix boasts that CompleteScan can ascertain 13 health factors, starting with heart rate and including blood pressure, body fat, weight ratios and the risk of stroke, type 2 diabetes and cardiovascular disease.
It reportedly can name a person’s emotion, a trick that many in the medical, mental health and AI communities have said is a speculative proposition. It has been claimed by a number of researchers recently, but the results have been controversial.
It brought controversy to NuraLogix, too, but of the political variety.
A lengthy piece in The Globe and Mail chronicles the company’s relationship with China’s Joyware Electronics, which the newspaper notes had ties to that nation’s autocratic government.
China has focused on facial recognition technology the same way that it invested in low-cost manufacturing a generation ago.
It has used a nationwide blanket of facial recognition systems to enforce Communist Party orthodoxy, political loyalty and anti-COVID restrictions. It is not a stretch to anticipate what the government might do with AI that it believes reveals dissent through emotion surveillance.
NuraLogix pulled back from the relationship. The Globe and Mail said that according to the company’s CEO, “no commercial products for public security resulted from the deal.”