Neural Networks for Keyword Spotting on IoT Devices

TL;DR

This paper investigates neural network architectures optimized for keyword spotting on IoT devices, focusing on reducing memory and computational requirements to enable efficient deployment on constrained hardware.

Contribution

The authors propose a CNN design specifically tailored for IoT devices that minimizes multiplies and model parameters, enhancing efficiency for keyword spotting tasks.

Findings

Achieved reduced multiply operations compared to standard models.

Designed a compact CNN suitable for low-resource IoT hardware.

Demonstrated effective keyword spotting performance on constrained devices.

Abstract

We explore Neural Networks (NNs) for keyword spotting (KWS) on IoT devices like smart speakers and wearables. Since we target to execute our NN on a constrained memory and computation footprint, we propose a CNN design that. (i) uses a limited number of multiplies. (ii) uses a limited number of model parameters.