A 174nW Event-Driven Keyword Spotting ASIC with DC-Drift-Resilient Dual-Mode Delta/SAR Quantizer and Multiplier-Less Processor
Recently, significant progress has been achieved in low-power keyword spotting (KWS) ASICs, which can serve as always-on wake-up function blocks for computationally intensive speech processing tasks. A 3-class CNN based KWS classifier consumed 510 nW power without including on-chip ADCs [1]. In [2], an ultra-low power event-driven level-crossing (LC) ADC was employed, reducing standby power to 148 nW. However, LC-ADC is vulnerable to DC drift. Other works include a 12-class KWS chip consuming $23 \mu \mathrm{~W}$ power [3], and a skip RNN-based 7-class KWS chip consuming 1.5 $\mu \mathrm{W}$ power [4]. Despite these technological advancements, researchers are still seeking KWS solutions with even lower power consumption, while not compromising on other key specifications such as keyword variety, classification accuracy and latency.
Authors
- Kaiji Liu
- Nan Wu (ORCID: https://orcid.org/0000-0002-1416-8862)
- Ning Pu (ORCID: https://orcid.org/0009-0001-7878-6408)
- Sining Pan (ORCID: https://orcid.org/0000-0002-5409-3367)
- Hanjun Jiang (ORCID: https://orcid.org/0000-0003-4911-0748)
- Zhihua Wang (ORCID: https://orcid.org/0000-0003-1732-6245)
- Yanshu Guo (ORCID: https://orcid.org/0000-0002-5800-8799)
- Siqi Zhang (ORCID: https://orcid.org/0000-0001-8533-2749)
- Yihuai Yang
Institutions
- Tsinghua University (CN)
Publication Details
- Published
- 2024-11-18
- DOI
- https://doi.org/10.1109/a-sscc60305.2024.10849242
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Science and Technology Major Project