Implementing Keyword Spotting on the MCUX947 Microcontroller with Integrated NPU

TL;DR

This paper demonstrates a real-time keyword spotting system on a microcontroller with an integrated NPU, achieving high accuracy and significant speedup, suitable for resource-constrained embedded devices.

Contribution

It introduces an optimized CNN-based KWS system with quantization on an embedded microcontroller with NPU, enabling efficient voice recognition on low-power devices.

Findings

59x inference speedup with NPU

97.06% accuracy achieved

Model size of 30.58 KB

Abstract

This paper presents a keyword spotting (KWS) system implemented on the NXP MCXN947 microcontroller with an integrated Neural Processing Unit (NPU), enabling real-time voice interaction on resource-constrained devices. The system combines MFCC feature extraction with a CNN classifier, optimized using Quantization Aware Training to reduce model size with minimal accuracy drop. Experimental results demonstrate a 59x speedup in inference time when leveraging the NPU compared to CPU-only execution, achieving 97.06% accuracy with a model size of 30.58 KB, demonstrating the feasibility of efficient, low-power voice interfaces on embedded platforms.