Ask any question about EPFL courses, lectures, exercises, research, news, etc. or try the example questions below.
DISCLAIMER: The Graph Chatbot is not programmed to provide explicit or categorical answers to your questions. Rather, it transforms your questions into API requests that are distributed across the various IT services officially administered by EPFL. Its purpose is solely to collect and recommend relevant references to content that you can explore to help you answer your questions.
Scaling always-on keyword spotting (KWS) system to larger vocabularies increases model size, driving higher memory and compute demands that push SoCs >100μW, with significant off-chip loading energy and area overhead. This work presents the first KWS SoC featuring a hybrid Cap-ROM-based in-memory computing (Cap-ROM-IMC)/5T-SRAM-based near-memory computing (SRAM-NMC) architecture with a transfer-learning based model supporting 20 keywords. Fabricated in 28 nm, the SoC achieves 92.1% accuracy on GSCD-v2 with the lowest reported power (1.16μW) and highest ROM memory density (2372.6KB/mm2) among prior arts.