A Cascade Architecture for Keyword Spotting on Mobile Devices

TL;DR

This paper introduces a cascade architecture for keyword spotting on mobile devices that combines low computational cost with DSP chips to enable continuous listening with minimal power use.

Contribution

The paper proposes a novel cascade architecture tailored for mobile keyword spotting that optimally balances accuracy and power efficiency.

Findings

Achieves low power consumption suitable for mobile devices.

Enables continuous listening for keywords.

Utilizes specialized DSP chips for efficiency.

Abstract

We present a cascade architecture for keyword spotting with speaker verification on mobile devices. By pairing a small computational footprint with specialized digital signal processing (DSP) chips, we are able to achieve low power consumption while continuously listening for a keyword.