IEEE CASS-EDS-SSCS & HUST Technical Talk No. 46 “ Ultra-low power AI accelerator design for always-on keyword spotting application”

#AI #accelerator #MFCC #ultra #low #power #design #KWS #ASIC
Share

As a widely used speech-triggered interface, deep-learning-based keyword spotting (KWS) chips are required for wake-up tasks. The always-on characteristic requires KWS to achieve ultra-low power while keeping good accuracy. This talk will show the design challenges and three case studies toward ultra-low power KWS chip design. First, a digital Keyword-Spotting Chip Using Serial-FFT-based MFCC and Binarized Depthwise Separable CNN, which for the first time pushed its power below 1μW.  Second, a sub-μW KWS chip with an acoustic activity detection (AAD) to achieve both low power and decent detection accuracy. Third, a full-system Near-Microphone Keyword-Spotting Chip, which integrated the analog front-end into the chip, along with optimized MFCC and Temporal Depthwise Separable CNN to achieve a full-system solution with only 608nW.


  Date and Time

  Location

  Hosts

  Registration


  • Add_To_Calendar_icon Add Event to Calendar
  • 1037 Luoyu Road, Wuhan, Hubei, P.R.China, 430074
  • Wuhan, Hubei
  • China 430074
  • Building: New building of SOEI
  • Room Number: Meeting room 117
  • Contact Event Hosts
  • Co-sponsored by Huazhong University of Science and Technology

  Speakers

Shan Weiwei

Biography:

WeiWei Shan (Member, IEEE) received the B.S. degree in microelectronics from Tianjin University, Tianjin, China, in 2003 and the Ph.D. degree in microelectronics from Tsinghua University, Beijing, China, in 2009. She was a Visiting Professor with Columbia University, New York, NY, USA, from 2017 to 2019. She is currently a Professor at the National ASIC Center, Southeast University, Nanjing, China. Her research mainly focuses on variation resilient, adaptive VLSI circuits, ultra-low-power SoC design, and countermeasure techniques of security circuits. She serves as technical committee in several international conferences such as IEEE CICC, A-SSCC, ICTA and so on. She has authored or co-authored over 50 technical articles in conferences and journals, including ISSCC, JSSC, IEEE TCASI/TCASII and TCAD.

Address:China