An ultra-low energy human activity recognition accelerator for wearable health applications

Ganapati Bhat, Yigit Tuncel, Sizhe An, Hyung Gyu Lee, Umit Y. Ogras

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Human activity recognition (HAR) has recently received significant attention due to its wide range of applications in health and activity monitoring. The nature of these applications requires mobile or wearable devices with limited battery capacity. User surveys show that charging requirement is one of the leading reasons for abandoning these devices. Hence, practical solutions must offer ultra-low power capabilities that enable operation on harvested energy. To address this need, we present the first fully integrated custom hardware accelerator (HAR engine) that consumes 22.4 μJ per operation using a commercial 65 nm technology. We present a complete solution that integrates all steps of HAR, i.e., reading the raw sensor data, generating features, and activity classification using a deep neural network (DNN). It achieves 95% accuracy in recognizing 8 common human activities while providing three orders of magnitude higher energy efficiency compared to existing solutions.

Original languageEnglish (US)
Article numbera49
JournalACM Transactions on Embedded Computing Systems
Volume18
Issue number5s
DOIs
StatePublished - Oct 2019

Fingerprint

Particle accelerators
Health
Energy efficiency
Engines
Hardware
Monitoring
Sensors
Deep neural networks

Keywords

  • Flexible hybrid electronics
  • Hardware accelerator
  • Health monitoring
  • Human activity recognition
  • Low-power design
  • Wearable electronics

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture

Cite this

An ultra-low energy human activity recognition accelerator for wearable health applications. / Bhat, Ganapati; Tuncel, Yigit; An, Sizhe; Lee, Hyung Gyu; Ogras, Umit Y.

In: ACM Transactions on Embedded Computing Systems, Vol. 18, No. 5s, a49, 10.2019.

Research output: Contribution to journalArticle

@article{85afb8b0745a4c05aeff01e6c0772066,
title = "An ultra-low energy human activity recognition accelerator for wearable health applications",
abstract = "Human activity recognition (HAR) has recently received significant attention due to its wide range of applications in health and activity monitoring. The nature of these applications requires mobile or wearable devices with limited battery capacity. User surveys show that charging requirement is one of the leading reasons for abandoning these devices. Hence, practical solutions must offer ultra-low power capabilities that enable operation on harvested energy. To address this need, we present the first fully integrated custom hardware accelerator (HAR engine) that consumes 22.4 μJ per operation using a commercial 65 nm technology. We present a complete solution that integrates all steps of HAR, i.e., reading the raw sensor data, generating features, and activity classification using a deep neural network (DNN). It achieves 95{\%} accuracy in recognizing 8 common human activities while providing three orders of magnitude higher energy efficiency compared to existing solutions.",
keywords = "Flexible hybrid electronics, Hardware accelerator, Health monitoring, Human activity recognition, Low-power design, Wearable electronics",
author = "Ganapati Bhat and Yigit Tuncel and Sizhe An and Lee, {Hyung Gyu} and Ogras, {Umit Y.}",
year = "2019",
month = "10",
doi = "10.1145/3358175",
language = "English (US)",
volume = "18",
journal = "ACM Transactions on Embedded Computing Systems",
issn = "1539-9087",
publisher = "Association for Computing Machinery (ACM)",
number = "5s",

}

TY - JOUR

T1 - An ultra-low energy human activity recognition accelerator for wearable health applications

AU - Bhat, Ganapati

AU - Tuncel, Yigit

AU - An, Sizhe

AU - Lee, Hyung Gyu

AU - Ogras, Umit Y.

PY - 2019/10

Y1 - 2019/10

N2 - Human activity recognition (HAR) has recently received significant attention due to its wide range of applications in health and activity monitoring. The nature of these applications requires mobile or wearable devices with limited battery capacity. User surveys show that charging requirement is one of the leading reasons for abandoning these devices. Hence, practical solutions must offer ultra-low power capabilities that enable operation on harvested energy. To address this need, we present the first fully integrated custom hardware accelerator (HAR engine) that consumes 22.4 μJ per operation using a commercial 65 nm technology. We present a complete solution that integrates all steps of HAR, i.e., reading the raw sensor data, generating features, and activity classification using a deep neural network (DNN). It achieves 95% accuracy in recognizing 8 common human activities while providing three orders of magnitude higher energy efficiency compared to existing solutions.

AB - Human activity recognition (HAR) has recently received significant attention due to its wide range of applications in health and activity monitoring. The nature of these applications requires mobile or wearable devices with limited battery capacity. User surveys show that charging requirement is one of the leading reasons for abandoning these devices. Hence, practical solutions must offer ultra-low power capabilities that enable operation on harvested energy. To address this need, we present the first fully integrated custom hardware accelerator (HAR engine) that consumes 22.4 μJ per operation using a commercial 65 nm technology. We present a complete solution that integrates all steps of HAR, i.e., reading the raw sensor data, generating features, and activity classification using a deep neural network (DNN). It achieves 95% accuracy in recognizing 8 common human activities while providing three orders of magnitude higher energy efficiency compared to existing solutions.

KW - Flexible hybrid electronics

KW - Hardware accelerator

KW - Health monitoring

KW - Human activity recognition

KW - Low-power design

KW - Wearable electronics

UR - http://www.scopus.com/inward/record.url?scp=85073160757&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85073160757&partnerID=8YFLogxK

U2 - 10.1145/3358175

DO - 10.1145/3358175

M3 - Article

AN - SCOPUS:85073160757

VL - 18

JO - ACM Transactions on Embedded Computing Systems

JF - ACM Transactions on Embedded Computing Systems

SN - 1539-9087

IS - 5s

M1 - a49

ER -