CEA-Leti Unveils Autonomous Imager

JOEL WILLIAMS, ASSOCIATE EDITOR
joel.williams@photonics.com

CEA-Leti has unveiled an autonomous imager technology that is able to activate smartphones and small appliances through facial recognition or other specific patterns, using 10,000× less power than similar technologies.

“The novel feature of the device is to allow a fully autonomous functioning of the image sensor, when in its surveillance mode,” CEA-Leti research director Antoine Dupret told Photonics Media. “Indeed, both the acquisition and the analysis of the scene are managed within the imager.”

The device is called μWAI (micro-WAY), and it’s about the size of a €1 coin, or a U.S. silver dollar. Its readout and processing architecture are co-designed with an algorithmic pipeline, in which the recognition results form a sequence of elementary algorithms to deliver ultralow power wakeup.

“In the surveillance ultralow power mode, the imager detects events within the scene,” Dupret said. “Upon the detection of a relevant event, the circuit first provides features for finer classification, then reverts into a standard image sensor functioning.”

Because the analysis occurs within the imager, rather than with the device’s primary computer system, power consumption is lower and the load on the computational system is reduced.

CEA-Leti's autonomous imager device µWAI. Courtesy of CEA-Leti.

The device features automatic exposure adjustments according to the ambient lighting conditions, and an 88-dB dynamic range, as well as motion detection and feature-extraction for event-based functioning, and AI-based object recognition.

It can be easily implemented into smart devices already using cameras, as the smart sensing part of the circuit can be implemented in most image sensors.

The technology behind μWAI will be introduced during CEA-Leti’s Leti Innovation Days, June 22-23.