Contact

Follow

  • Black Twitter Icon

(US) +1-202-468-8467

(International) +1-408-306-0658

Palo Alto, CA | Burlington, MA | Washington, DC

United States

©2018 BY CAMVI TECHNOLOGIES | ALL RIGHTS RESERVED

Biometrics on the Edge: AI-Powered Solutions for Next Generation Biometric Systems

Presented by Matt McCrann during the Law Enforcement & Biometrics session at FedID 18, Tampa FL 

 

The next generation of biometric requirements for government users will include the continued expanding scope of biometric collection, improving the efficiency of existing systems, and providing more actionable and timely intelligence for officers, agents, and soldiers.

 

These growing requirements create the need to take the legacy biometric systems and databases that were primarily tethered to server rooms and data centers due to processing constraints, and move these capabilities further out to “the edge” of operation, or in the world of biometrics, closer to the point of collection or enrollment. Biometrics on the Edge will be the culmination of this effort; the decentralization of biometric capture, processing, and dissemination of actionable intelligence across the entire operational network.

 

In order to achieve the operational flexibility that decentralized solutions bring, these next generation solutions will need to be developed with three core competencies; these systems need to be highly scalable, interoperable, and adaptable.

 

Scalability

 

Scalable systems provide a consistent capability and user experience, even as the scope of the operation, system, or sub-component, such as the database size or number of users, increases or decreases.

 

Today, the continued increase in data that these systems are expected to handle come from two main directions; the increase in known data, such as reference databases, and the increasing amount of incoming, often real-time data from sensors and other collection devices such as cameras.

 

The increasing size of databases demands consideration for processing power. More data, more information to search, means more CPUs will be required to produce results in a timely manner. Just as the size of database(s) directly impacts processing requirements, the increase in sensors and volume of new, incoming data will directly impact bandwidth requirements. As a result, the development and deployment of next gen systems will need to address both processing power and bandwidth concerns.

 

Interoperability

 

Interoperable systems effectively share critical information or intelligence with other systems and personnel. The focus on interoperability should be the seamless sharing of this information with the systems already deployed and with the appropriate personnel through their preferred, existing channels for new information.

 

Standards for biometric collection, which provide a common language and data format, can greatly increase the interoperability of disparate systems. Standards have long existed and served well for fingerprints, so as other modalities rise in popularity, it will serve governments well to adopt common standards across agencies to support face, iris, palm, gait and other biometrics of interest.

 

From the industry perspective, providing application programming interfaces (APIs) and support for common and open data formats, will allow new biometric systems to more easily integrate with 3rd party systems and the larger operational picture.

 

Adaptability

 

The benefit of a system’s ability to adapt to a quickly changing operational environment is apparent. No department wants to go through the process of completely replacing its solutions every 18-24 months because they no longer address the needs of the mission.

 

Providing software and hardware, agnostic to as many other components as possible, is an advantageous way for Industry to maximize their solutions’ adaptability. Offering decentralized solutions that can both function fully standalone and plug into the larger network of operations when necessary, should be a prime objective during development.

 

AI: Enabler of Next Generation Biometric Systems

 

Artificial Intelligence is more substantial than its current buzzword status, however, it still needs to be demystified.

 

The term Artificial Intelligence was first defined by John McCarthy in 1956 as “The science and engineering of making intelligent machines.” Since then, all subsequent definitions of AI pair the words “intelligent” and “machines,” “computers,” or “agents.” But what does that actually look like for biometrics in an operational sense? What value does an “intelligent machine” actually provide end users as it relates to facial recognition, voice, and other biometric modalities?

 

AI systems will always demonstrate two things; a level of autonomy and form of adaptiveness.

 

Autonomy means that the machine, system, solution will be able to make a decision; i.e. return a result, without a predefined answer. This rules out all defined functions and spreadsheet formulas, no matter how complex. An example of this sort of autonomy would be the matching

of an unknown or “wild” biometric sample to a known collection, without human interaction or verification. But a level of autonomy alone is not indicative of an AI-powered system.

 

Adaptiveness is the system’s demonstrated ability to respond to circumstance or change in environment. An example of adaptiveness for a biometrics is the increased accuracy in the systems facial recognition capability as more data, in this case face images becomes available.

 

The biggest advantages to fielding autonomous and adaptive next generation systems are the systems’ ability to overcome real world challenges. Adapting to disparities in biometric data collected by multiple systems, substandard quality, low resolution, or limited number of data points are all examples of challenges that AI-powered solutions can overcome. AI solutions can be trained to mitigate inconsistencies, obstructions, atmospherics, and differing angles in order to acquire and match based on a biometric from video such as face, iris, or even fingerprint. With AI, substandard or throw-away data can be used in training models that can result in previously unusable data becoming actionable intelligence.

 

When next generation systems leverage AI, and more specifically, deep learning, this serves as a catalyst for highly scalable, interoperable, and adaptable solutions, capable of addressing the most challenging problems in biometrics.

 

Where We Go from Here

 

The next wave of biometric solutions should be built with scalability, interoperability, and adaptability as key considerations. Solutions offering these core competencies will provide a cohesive capability and consistent user experience throughout the operational network.

 

Standards can serve as a primary driver for interoperability and decentralized solutions will lead to putting more adaptable solutions in the hands of end users. By leveraging Artificial Intelligence and deep learning, these next generation systems will be able to adapt more quickly to changes in mission objective, environment, and the challenges end users will face with an ever growing amount of available data.

 

Related Camvi News (with Video)

 

Camvi Spoke at FedID on Next Generation Biometrics on the Edge

FindBiometrics: Camvi Highlights Importance of AI in Biometrics

Share on Facebook
Share on Twitter
Share on LinkedIn
Please reload