Congratulations to Sensors Group students Sheng Zhou and Zixiao Li, also ex-member Kwantae Kim ( now Ass Prof. at Aalto University), and Tobi Delbruck and Shih-Chii Liu for the Best ISSCC Demonstration Session Award for their neuromorphic spoken language understanding chip called FENNEC (named after fox with really big ears).
FENNEC recognizes about 200 phrases like “Change the language to German” or “Turn on the lights in the bedroom” and maps them to 32 (action,object,location) outputs with real-world accuracy >90%. FENNEC burns <10uW and is only 2mm2 in 65nm process. The 16-channel cochlea analog front end enables high-accuracy over large speaker distances with global and channel automatic gain control. The digital gated recurrent network uses temporal delta sparsity to reduce power to nearly zero during quiet times.
Further Information:
ISSCC Award Website: https://static1.squarespace.com/static/6130ef779c7a2574bd4b8888/t/699b3b4225785065925a9263/1771780930299/ISSCC2026-AwardsHandout.pdf#page=5.61
FENNEC Project Website: https://sensorsini.github.io/fennec/
FENNEC Paper: https://ieeexplore.ieee.org/document/10904788
FENNEC Demo Video: https://www.youtube.com/watch?v=GLxwnimXB2U