Outnumbering its previous year patents, in 2019, Amazon has secured around 2,396 patents. While many patents like robotic arms that toss warehouse items and real-time accent translator, this year it’s a non-contact biometric identification system. The application for the patent was granted on 26 December 2019, which states that the technology includes a hand scanner that generates images of user’s palm for identification. The idea is to replace the use of application or car to pay at Amazon Go and Whole Foods for quickly authenticating and checking out.
Technology Behind
The whole system comprises an infrared light source, a camera, and a controller. The technology utilises computer vision to take images of users’ hand and leverage the details of biometric, wrinkles, and veins for identifying users quickly. For this, it emits two polarised lights to collect the details of the users’ hand. This allows the camera to acquire two sets of raw images for further processing and depicting human palm in a canonical orientation, canonical pose, and vein patterns with the controller.
A neural network is also used for characterising the surface of a human palm, and another neural network is deployed for describing internal anatomical structures of the human palm. While the first neural network throughputs feature vector expressed in an n-dimensional space, the second one outputs feature vector expressed in a p-dimensional space, where p is less than n. These outputs are fed to a third neural network that generates a signature which is used to match with the reference signature and determine user identity for quickly authenticating people.
Use Cases Of Non-Contact Biometric Identification
As per the patent filing, the technology can be used in stores, libraries, hospitals, and offices. People can hover over the scanner where they get their palm snapped by the camera for authentication, instead of using a card or a mobile application. Currently, at Amazon Go, buyers use a mobile application to validate and then shop at the store. However, with this new technology, purchasers in the future will be able to instantly verify their identity without the need for any physical devices or cards. The technique can be used to associated with users’ accounts that can be used for the payment as soon as a user moves out of the store.
This will be another endeavour for Amazon in its mission of simplifying the shopping experience for users. Amazon’s patent for one-click online ordering enabled it to gain an edge over other buyers, and now with the non-contact biometric identification system, it will allow users to shop effortlessly.
Apart from shopping at the stores, the technology can be used at material handling facility, office transportation facility, or other locations that require controlled based identity.
Potential Risk
Privacy-related concerns are on the rise among users, thus, Amazon needs to ensure people trust in their new technology. Amazon is already facing criticism for its security lapses in its camera technology, which was hacked and spooked the kids at home. Unlike passwords that can be changed in case of a hack, however, biometrics details cannot be altered, resulting in a permanent theft of identity. However, Amazon rejected the privacy issues concerns by saying that the technology is trustworthy than swiping a credit card. Besides, it also said, the strenuous nature of biometric characterisation is impossible to copy or transfer.
Outlook
Amazon with non-contact biometric identification system can shift the way validation is carried out and has the potential to replace facial recognition technology in a few use cases. Facial recognition technology has been used for gaining access control and accrediting user identity but has witnessed backlash among users, thus for the new non-contact technology can be equipped to mitigate the problems associated with surveillance.
However, not many patented technologies have actually been introduced to the market, thus, we will have to see if Amazon can materialise the idea into the real-world use cases.
