Biometrics are increasingly used for identification and mobile payments, but most current systems face limitations and concerns about privacy. Maybe you’ve worried about your facial data being held by various tech companies or hesitated about pressing your finger on a dirty scanner at border checks.

Touchless palm verification is one piece of tech that might be able to address these problems. Palm recognition systems work by identifying vein patterns and lines and creases on the hand’s surface, ideally using cameras and infrared to avoid contact.

“Compared with [a] face, palmprint is not privacy sensitive,” said Xu Liang, who has studied palm recognition and works at Chinese University of Hong Kong’s Shenzhen Institute of Artificial Intelligence and Robotics for Society. “Given a palmprint image, no one could tell to whom it belongs. In daily life, the palm is curled up, so without active cooperation, it’s very hard for hidden cameras to steal our palmprint information.”

Facial recognition has become ubiquitous in China, but it’s recently been plagued by privacy concerns. (Picutre: Li Jianan/Xinhua)

Liang was a research assistant for Professor David Zhang, the founding director of Biometrics Research Center at the Hong Kong Polytechnic University. Zhang has also studied palm recognition for years, publishing many works related to the field since 2003.

Another security advantage for palm recognition is that it’s difficult to steal when the target is dead. Liang said subcutaneous palm vein patterns can only be captured under infrared light when the subject is alive, giving the system an inherent anti-spoofing ability.

Some startups in China have already started bringing this technology to market, although not on a large scale.

DeepBlue Technology, a 5-year-old AI startup, has several palm recognition products for door access and vending machines that support palm recognition payments. Melux, another startup, has a palm recognition product called AirWave, which the company says has been deployed in some office areas in Guangzhou.

I hoped to learn more about the development of these products, but no one from DeepBlue was available to comment before Chinese New Year. Melux didn’t respond to a request for comment.

American tech companies are working on palm recognition, too.

Recode reported in December that Amazon filed for a patent on a touchless scanning system that identifies people by their palms. The application says that the system captures images of both surface characteristics like wrinkles and deeper characteristics such as veins. Reports from September suggested that the company was planning a hand payment system for Whole Foods.

Apple also has a palm verification patent of its own, which was filed last January and published by the US patent office in September.

Apple’s patent application for a “palm biometric sensor layer” suggests that it could be used on devices like phones and watches. (Picture: United States Patent and Trademark Office)

Many patented technologies never see the light of day. But even beyond security concerns, Liang said palm recognition has a variety of advantages over facial and fingerprint recognition.

Fingerprint identification can’t be used when fingers are wet and is unusable for certain people like construction workers, whose fingerprints could be worn out from daily work, he said. And a stained contact surface after being touched many times can also lead to unstable imaging. 

While facial recognition also doesn’t require contact, it poses security risks because face images are easy to steal, and many people are resistant to it over privacy concerns. While China has been aggressively deploying facial recognition, facial data theft is rampant in the country. In December, state broadcaster CCTV reported that pictures of more than 5,000 faces were being sold online for less than US$2.

In China, 10 yuan (US$1.42) could buy you 5,000 faces. (Picture: CCTV)

But even though palm recognition tech exists today, it might not be ready for widespread use.

Last year’s LG G8 ThinQ smartphone, for instance, includes a “Hand ID” feature. It uses a Time of Flight (ToF) sensor to detect depth and an infrared emitter to map out and and verify the vein layout of a user’s palm.

Theoretically, users need only to hover a hand above the phone as it rests on a table to unlock it. In reality, though, reviews said Hand ID was “slow to detect a match” and “a total pain to deal with” because it was difficult to hover the palm where the sensors needed it to be.

Hand ID on the LG G8 ThinkQ “uses your unique circulatory and hand patterns to effortlessly unlock your phone with a quick scan of your palm.” In theory, at least. (Picture: LG)

Another issue, according to Liang, is that palm recognition is more complicated and more expensive to make.

“Its requirement for device development is a little higher than face and fingerprint identification systems,” Liang said. “I think this is a reason why palmprint identification hasn’t been as popular.”

The reason is that palms have finer details than faces and a larger area to identify than fingers, he said. Images also need to be captured at a specific distance with good lighting. That requires customized imaging modules and deep cooperation with camera module manufacturers.

Another reason it’s not as popular is that there aren’t any large-scale palmprint datasets available, according to Liang.

“For handling different palm shapes and postures, a large-scale palmprint database is urgently needed,” he said. “In this age, all smart systems are driven by big data.”

Liang also said that citizens in China all have to register their face and fingerprints with the police, and they’ve gotten used to face and fingerprint recognition systems at border checks. The government doesn’t have such support for palm-based systems, so there isn’t the necessary publicity and capital investment for palm recognition to take off. 

But there could still be a future for it.

“I think the situation is getting better since more and more companies and organizations have paid attention to palmprint identification,” Liang said.