Answers are grounded in this paper and cite their sources.
This article presents a fully analog, signal-to-noise-ratio-based voice-activity detection (VAD) circuit. The circuit is based on an energy-efficient analog implementation with continuous-time nonlinear operation and fully passive switched-capacitor signal processing. The overall chain is composed of a programmable-gain amplifier, a squarer, an integrator, an SC-based signal averaging circuit, and a periodic threshold update circuit for adaptability. This implementation allows the minimization of both power consumption and chip area. The VAD circuit prototype has been fabricated in a 180-nm CMOS technology and occupies an area of 0.14 mm 2 . The device achieves 99.5% classification accuracy in a domestic environment in the presence of loud ambient noise, consuming 760 nW from a 1.2-V supply.
No takes yet. Share an insight, caveat, or question.
Croce et al. (2020) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: