The increasing diffusion of tiny wearable devices and, at the same time, the advent of machine learning techniques that can perform sophisticated inference, represent a valuable opportunity for the development of pervasive computing applications. Moreover, pushing inference on edge devices can in principle improve application responsiveness, reduce energy consumption and mitigate privacy and security issues. However, devices with small size and low-power consumption and factor form, like those dedicated to wearable platforms, pose strict computational, memory, and energy requirements which result in challenging issues to be addressed by designers. The main purpose of this study is to empirically explore this trade-off through the characterization of memory usage, energy consumption, and execution time needed by different types of neural networks (namely multilayer and convolutional neural networks) trained for human activity recognition on board of a typical low-power wearable device. Through extensive experimental results, obtained on a public human activity recognition dataset, we derive Pareto curves that demonstrate the possibility of achieving a 4× reduction in memory usage and a 36× reduction in energy consumption, at fixed accuracy levels, for a multilayer Perceptron network with respect to more sophisticated convolution network models

Exploring Artificial Neural Networks Efficiency in Tiny Wearable Devices for Human Activity Recognition

Lattanzi, Emanuele
;
Freschi, Valerio
2022

Abstract

The increasing diffusion of tiny wearable devices and, at the same time, the advent of machine learning techniques that can perform sophisticated inference, represent a valuable opportunity for the development of pervasive computing applications. Moreover, pushing inference on edge devices can in principle improve application responsiveness, reduce energy consumption and mitigate privacy and security issues. However, devices with small size and low-power consumption and factor form, like those dedicated to wearable platforms, pose strict computational, memory, and energy requirements which result in challenging issues to be addressed by designers. The main purpose of this study is to empirically explore this trade-off through the characterization of memory usage, energy consumption, and execution time needed by different types of neural networks (namely multilayer and convolutional neural networks) trained for human activity recognition on board of a typical low-power wearable device. Through extensive experimental results, obtained on a public human activity recognition dataset, we derive Pareto curves that demonstrate the possibility of achieving a 4× reduction in memory usage and a 36× reduction in energy consumption, at fixed accuracy levels, for a multilayer Perceptron network with respect to more sophisticated convolution network models
File in questo prodotto:
File Dimensione Formato  
sensors-22-02637.pdf

accesso aperto

Tipologia: Versione editoriale
Licenza: Creative commons
Dimensione 954.81 kB
Formato Adobe PDF
954.81 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11576/2698009
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 25
  • ???jsp.display-item.citation.isi??? 21
social impact