Gait over Here! Will China’s Machine Vision and AI Startups Help Gait Biometric Recognition Back on Its Feet?

High-value, innovative startups from China have recently acquired their fair share of the spotlight when it comes to surveillance, biometrics, and artificial intelligence. SenseTimeis an AI and machine vision company that also focuses on highly advanced face recognition, liveness detection, feature points, and attributes and was dubbed a unicorn even before it left stealth mode (amassing, in total, more than US$2.5 billion in funding). Megviiis another AI company that focuses almost exclusively on face/image/object recognition, video, and behavioral analytics (with an impressive portfolio for surveillance system solutions), securing more than half a billion dollars in funding and a recent evaluation of US$23.5 billion. Both firms are backed by the Alibaba Group, which is steadily establishing its presence in the AI arena.

Joining the ranks of innovative Chinese startups is Watrix, a company that on top of intelligent transportation and face recognition also focuses on a somewhat forgotten modality: gait biometrics. Gait is perhaps the least known biometric modality (comparable perhaps only to hand geometry) and has failed to attract attention in recent years. As AI, biometrics, machine vision, object, and movement recognition technologies evolve, however, its value proposition becomes even more relevant with each passing year and it looks like new companies like Watrix can help it get back on track.

Gait-based biometric identification (i.e., making use of movement, body, and pressure pattern data to identify individuals) has been part of behavioral biometric technologies for nearly two decades. Although other behavioral and non-physiological modalities like voice recognition, online user behavior, or keystroke dynamics have enjoyed a plethora of implementations over the last years (albeit with somewhat less than satisfactory results in some cases), gait recognition has definitely lagged behind. FST Biometrics, an Israeli firm that made use of a combination of gait and face recognition for military and security checkpoints, has not managed to attract much attention over the last years. Until very recently, most of the worthwhile gait recognition discourse originated from academic environments and IEEE journal publications, never having the chance to enjoy the spotlight that other modalities did in the past.

The reason for the above is threefold:

• First, the lack of accuracy for certain applications that have higher standards in security.
• Second, the significant technological upgrade required in some cases (e.g. motion, weight, pressure sensor plates) for a result that can be achieved with a much more versatile implementation using other modalities.
• Third, the lack of applications worthy enough to gather the required attention from more implementers.

As one would expect, gait recognition is not easily employed in any consumer, commercial, or enterprise settings or even adjacent verticals. In addition, the lack of accuracy and the infrastructure investment required for high-security for some gait applications are not the most compelling points currently at the top of the budget list.

The gait modality, however, is due for a return. Many innovative applications have been suggested, including pressure sensitive plates in combination with 3D cameras for border control applications or door, rear, and side-view cameras for remote engine ignition for the automotive market. Not all of them will reach critical mass, especially when other more reliable modalities are higher up in the security value chain and currently awaiting implementation like iris, face, and voice recognition. Technology-wise, it would also make sense to make use of much more reliable applications like NIR (Near Infrared Spectroscopy), utilized in iris recognition, or further invest in 3D imaging liveness detection for face recognition, which is used in a much wider spectrum of applications across different verticals.

The sweet spot for gait recognition, however, is exactly the one incorporated in the mission statements of many of the new innovative Chinese startups: advanced video and behavioral analytics for surveillance and security purposes. Renewed by adjacency technology effects, ABI Research posits that the biometrics, security, and surveillance markets will steadily begin to reconsider the use of gait biometrics in the long term.

The adjacent technologies that are expected to assist gait in making its next crucial step are AI, deep learning, automation, video analytics, and machine vision. Watrix’s gait recognition system is already piloting with Chinese authorities and employs new machine learning algorithms to determine people’s identity based on their movement, silhouettes, pace, and other behavioral patterns in real-time without the use of any face recognition. The recent surveillance surge originating in APAC and instigated primarily by the Chinese government will further enhance the transition of gait recognition toward existing machine vision and video analytics ventures. However, the future will tell if gait recognition can be properly introduced into “mainstream” biometrics applications.