After the iWatch became all the rage, wearable devices have suddenly become the most popular IT product. Companies in the market rushed to introduce various devices, such as smart watches, smart wristbands, and even smart necklaces and earrings, all of which can detect biosignals through built-in sensors. However, most wearable devices share common drawbacks. In addition to inaccurate detection under vigorous exercises like jogging, some medical devices require users to change their lifestyle, which significantly reduced their use rate. SmartFace Inc, led by professor Bing-Fei Wu of National Chiao Tong University (NCTU), has developed a vision-based biosignal health management system to measure and detect biosignals through visuals without contact. The system's accuracy is higher than that required by medical standards and can be applied in medical and financial sectors.
The application of wearable devices in the medical sector started early, then the IT industry launched related products, including all kinds of smart wristbands, glasses, and watches. These wearable devices use built-in sensors to detect biosignals, but they lose their accuracy if not aiming at certain parts of the body (i.e. pulse). As for medical grade products that require a high standard on precision, some need to be worn over long periods of time, such as heart rate detectors worn by cardiac disease patients at home or devices for detecting sleep apnea. These devices affect users' daily living. In light of this, Wu focused on precision and easy application as the two major principles when conducting research and development. His vision-based biosignal health management system can accurately detect biological information without users feeling anything.
This system first obtains facial information through a camera and then shows the heart rate and blood pressure after the analysis by special algorithms developed by FaceHeart Inc. Accuracy can reach 2-3 bpm when users are inactive, not worse than the medical standards. It can still be use even during jogging on the treadmill. Also, users don't need to purchase a high-precision camera. Heart rate detection only requires cameras with a frame speed of 30 fps, whereas the detection of blood pressure will require cameras with higher specs, but Taiwan suppliers are capable of making both kinds of cameras.
The vision-based biosignal health management system of FaceHeart Inc. has diverse applications. It is for users who need to detect biosignals without making contact with the device itself. The system is most commonly seen in elderly care. To take care of elders in the family, some people will purchase devices like smartwatches for them to wear, but the elderly are not familiar with electronic products, which are eventually cast aside without being used. The system from FaceHeart Inc. can have a camera set up in front of the television so the elderly can have their heart rate and blood pressure detected while watching TV.
In addition to the medical sector, financial institutions are starting to use the vision-based biosignal health management system. In 2017, Professor Wu's team at NCTU won first prize in the smart financial system competition by the Shanghai Commercial and Savings Bank, Ltd. (SCSB). They then worked with SCSB to integrate it into the Know Your Customer (KYC) system, which companies use to identify customers. Scams are on the rise in recent years. In the past KYC tried to prevent dummy accounts by relying on questionnaires to conduct background checks on customers applying to open new accounts. But SCSB uses the vision-based biosignal identification system, allowing employees to immediately determine whether the applicants passes the KYC procedures, reducing the employees' workload.
Both the hardware and software of this health management system are independently researched and developed by Wu's team at NCTU. Even the microprocessing chip comes from the cooperation with Taiwan's MediaTek Inc. It is an AI system that is completely self-made in Taiwan. Professor Wu pointed out that the system is extremely convenient to use, but they encountered many challenges during research and development. For example, they need to figure out how to ensure accurate measurement under circumstances without ample light, such as nighttime or in a dark room. Also, cameras on the market automatically adjust the aperture and shutter. This image also affects the analysis of AI. The R&D team spent 5 years overcoming all challenges and smoothly commercialize this new technology.
As for the business model, Wu says that technology will have two directions: One is to authorize the core software to companies so the system can be built into cameras. The FaceHeart Inc. team can customize the software to suit customer products. The other direction is to partner with existing camera companies. For instance, cameras from security monitoring companies can combine FaceHeart's Al system to have more added functions. Wu expressed that incorporating AI into visuals has become a trend in the industry, but FaceHeart does not delve into mass facial recognition but a rather accurate biosignal identification to differentiate its market from competitors and aim for high-value business opportunities in a new territory.