FANNCortexM: An Open Source Toolkit for Deployment of Multi-layer Neural Networks on ARM Cortex-M Family Microcontrollers : Performance Analysis with Stress Detection

Abstract

We present FANNCortexM, an open-source toolkit built upon the Fast Artificial Neural Network (FANN) library to run lightweight neural networks on ARM Cortex-M series microcontrollers. The toolkit takes a neural network trained with FANN and generates code targeted at execution on low-power microcontrollers either with a floating-point unit (i.e., ARM Cortex-M4F and M7F) or without a floating-point unit (i.e., ARM Cortex M0-M3). The toolkit is optimized in terms of memory and computational resources. We demonstrate its functionality on the basis of a sample application scenario performing stress detection on a wearable multi-sensor bracelet. Experimental results show a high classification accuracy of 96% for the target application scenario, and low latency of only a few microseconds while keeping the memory requirements (11kB flash storage, 36kB RAM) far below the limitations of the device. Power measurements show a power consumption of only 1.6mW, thus allowing continuous stress detection for 29 days.