A 760 nW, 180 nm CMOS Analog Voice Activity Detection System

Abstract

This paper presents a fully analog, signal-to-noise ratio (SNR) based voice-activity detection circuit, which 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. The circuit exploits an energy-efficient analog implementation with continuous-time non-linear operation and fully-passive switched-capacitor processing, to minimize both the power consumption and the chip area. The VAD circuit prototype, fabricated in a 180 nm CMOS technology, occupies 0.14mm2.