You are here : Home > INDUSTRIAL INNOVATION > DEMOS > Unlocking devices upon face recognition for a few additional micro Watts

Articles & files | Focus | Article


µWAI



Published on 25 January 2021


Unlocking devices upon face recognition for a few additional micro Watts  


   What is µWAI?

CEA-Leti's autonomous imager µWAI activates any device upon recognition of a specific pattern — e.g., unlocking a smartphone upon the recognition of a face.

It is the first highly efficient, compact and ultra-low power smart awaken system designed for everyday small appliances:

  • Consuming 10,000 × less than low-power camera plus processor set
  • Highly compact: 1 € coin
  • Privacy-compliant AI-based recognition: near human detection performance (95%)
  • wide operating lightening condition
  • "Always-on" CR1025 Battery lifetime: 5 years
  • 3-6µW operation  

 


  Applications:

  • Mobile devices: automatic switching, face identification.
  • Smart home: contact-less smart switches of household appliances, sport and entertainment devices.
  • Smart building/smart home: face recognition, people counting, alarm triggering
  • Automotive: in cabin situation awareness, driver identification   
  • Automotive: out cabin parking situation awareness, smart door unlocking system



   What's new? :


µWAI features a novel readout and processing architecture co-designed with an optimized algorithmic pipeline providing ultra-low power wake-up modes and compact silicon implementation. It is the first smart image sensor jointly featuring:

  • Auto-exposure for all lightening condition and 88dB dynamic range,
  • Motion detection and feature extraction for event based functioning,
  • AI-based object recognition triggering highly reliable identification.

These key features enable highly reliable decision taking for few tens of pJ/pixel/frame, outperforming the best "off-the-shelf" solutions.



  What's next?

  • The core of CEA-Leti's autonomous imager is its recognition engine. CEA-Leti's team is working hand in hand with an industrial partner towards the development of specific smart imager products.

  • This technology can be applied to other use cases, leveraging the institute expertise in integrating a compact recognition engine within a smart image sensor. CEA-Leti tailors competitive solutions to meet the requirements of its partners.


KEY FACTS

  • Presentation at the highly selective VLSI 2020 conference: A. Verdant et al., "A 3.0μW@5fps QQVGA Self-Controlled Wake-Up Imager with On-Chip Motion Detection, Auto-Exposure and Object Recognition," 2020 IEEE Symposium on VLSI Circuits, Honolulu, HI, USA, 2020
  • 10 patents


  •      Flyer

             (406 Ko)