CEA-Leti Develops Novel Architecture for Keyword-Spotting (KWS) In Always-On, Voice-Activated Edge-AI Systems
ISSCC 2024 Presentation Reports Accurate Speech Recognition At Power Consumption Below One Microwatt
GRENOBLE, France – February 22, 2024 – CEA-Leti has developed a keyword-spotting system that dramatically improves accuracy in always-on, voice-activated Edge-AI systems and that consumes less power in a far smaller silicon footprint than current technology.
Presented in a paper at ISSCC 2024 in San Francisco, the new architecture uses time-domain signal processing on oscillators locked by injection and is suitable for devices running on energy harvesters, which supply power below 0.5V. The paper, “0.4V 988nW Time-Domain Audio Feature Extraction for Keyword Spotting Using Injection-Locked Oscillators”, reports accurate speech recognition at power consumption below one microwatt.
It describes the first injection-locked, oscillator-based time-domain audio feature extraction (TD-FEx) demonstrating keyword spotting operating down to 0.4V, while achieving 91 percent accuracy on 10 words. TD-FEx information is not coded as a voltage but as a time delay of two clocks’ signals. In addition to being well suited for advanced nodes, its advantages are digital-like implementation with low-supply voltage and better noise immunity than current systems. CEA-Leti’s system demonstrated accurate speech recognition with power consumption below 1 µW.
Some analog-based audio feature extraction (FEx) units using multi-channel Gm-C bandpass filters can supply 10 times the power efficiency of digital FEx units in a comparable silicon area. “However, analog FEx circuits have not demonstrated KWS with more than four keywords,” the paper reports. “They also suffer from a large footprint, challenging technology migration and limited dynamic range at low supply voltage, while speech signals have inherently a high dynamic range.”
“Our system’s silicon area of 0.15mm2 is at least 3.5 times smaller than prior art on the same process node of 65nm,” said Ali Mostafa, lead author of the paper. “With a power of 988nW, our system is nine times more power-and-area efficient than ring-oscillator-based TD-FEx.”
Applications beyond speech recognition for this system include predictive maintenance and health monitoring that require on-line frequency decomposition of the sensor data.
About CEA-Leti (France)
CEA-Leti, a technology research institute at CEA, is a global leader in miniaturization technologies enabling smart, energy-efficient and secure solutions for industry. Founded in 1967, CEA-Leti pioneers micro-& nanotechnologies, tailoring differentiating applicative solutions for global companies, SMEs and startups. CEA-Leti tackles critical challenges in healthcare, energy and digital migration. From sensors to data processing and computing solutions, CEA-Leti’s multidisciplinary teams deliver solid expertise, leveraging world-class pre-industrialization facilities. With a staff of more than 2,000 talents, a portfolio of 3,200 patents, 11,000 sq. meters of cleanroom space and a clear IP policy, the institute is based in Grenoble, France, and has offices in Silicon Valley, Brussels and Tokyo. CEA-Leti has launched 75 startups and is a member of the Carnot Institutes network. Follow us on www.leti-cea.com
|
Related News
- CEA-Leti & Stanford Target Edge-AI Apps with Breakthrough NVM Memory Cell
- CEA-Leti Launches OpenTRNG, an Open-Source Project For True Random Number Generators Using Ring-Oscillator-Based Architectures
- CEA-Leti Announces Launch of FAMES Pilot Line As Part of EU Chips Act Initiative
- CEA-Leti Paper in Nature Communications Reports First Complete Memristor-Based Bayesian Neural Network Implementation For Real-World Task
- CEA-LETI Develops Circuits for Neuromorphic Processors That Replace CMOS Transistor-Based Tcam Memory With Rram-Based Tcam Memory
Breaking News
- Silicon industry veteran Oreste Donzella joins Sondrel board as Non-Executive Director
- Powering the NVM and Embedded Chip Security Technologies
- BOS and Tenstorrent Unveil Eagle-N, Industry's First Automotive AI Accelerator Chiplet SoC
- BBright Expands Ultra HD Capabilities with intoPIX JPEG XS Technology in its V2.2 Decoder Platform
- Jmem Tek and Andes Technology Partner on the World' s First Quantum-Secure RISC-V Chip
Most Popular
- MosChip selects Cadence tools for the design of HPC Processor “AUM” for C-DAC
- Cadence and Rapidus Collaborate on Leading-Edge 2nm Semiconductor Solutions for AI and HPC Applications
- Quobly announces key milestone for fault-tolerant quantum computing
- Synopsys Announces Industry's First Ultra Ethernet and UALink IP Solutions to Connect Massive AI Accelerator Clusters
- Alphawave IP - Announcement regarding leadership transition
E-mail This Article | Printer-Friendly Page |