HAR-ASIC: A Configurable Edge-AI Human Activity Recognition Circuit for Ultra-Low Power Wearable Devices
Open Access
- 30 September 2026
- journal article
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 10 (3) , 1-24
Abstract
We present an application-specific integrated circuit (ASIC) using digital hardware design to maximise energy-efficiency of human activity recognition (HAR) in autonomous wearable devices. The HAR-ASIC was designed in 22 nm technology and includes a battery of configurable, hardware-optimised feature extraction functions along with a decision tree ensemble classifier, thus serving as an edge AI device for local sensor data processing. We validate the HAR-ASIC based on post-layout process data with five HAR datasets and show the design scalability for configurations to maximise energy saving and others to maximise recognition performance. We show that feature extraction often requires up to 100 times more energy than the classification function. Our configuration analysis shows that feature extraction energy consumption can be reduced by factors of two to five, with at most 3% loss in F1-score, depending on the dataset. Across all datasets, feature extraction and classification energy was between 64 and 678 nJ per inference for Pareto-optimal configurations using sliding window buffers of up to 512 samples. We conclude that the highly configurable HAR-ASIC design could be the basis to realise ultra-low power sensor-edge AI systems for continuous wearable HAR and possibly further edge AI applications.Keywords
Funding Information
- German Federal Ministry for Research, Technology and Space (16ME0382K)
- University of Freiburg (Open Access Fund)
This publication has 28 references indexed in Scilit:
- An Experimental Study on the Energy Efficiency of Feature Selection for Human Activity Recognition with Wrist-Worn Devices2025
- Energy-aware human activity recognition for wearable devices: A comprehensive reviewPervasive and Mobile Computing, 2024
- A multi-channel hybrid deep learning framework for multi-sensor fusion enabled human activity recognitionAlexandria Engineering Journal, 2024
- A comprehensive comparison of machine learning approaches with hyper-parameter tuning for smartphone sensor-based human activity recognitionMeasurement: Sensors, 2023
- A Survey on Human Activity Recognition Using Deep Learning Techniques and Wearable Sensor Data2022
- Energy-efficient activity recognition framework using wearable accelerometersJournal of Network and Computer Applications, 2020
- Improving physical activity recognition using a new deep learning architecture and post-processing techniquesEngineering Applications of Artificial Intelligence, 2020
- LSTM-CNN Architecture for Human Activity RecognitionIEEE Access, 2020
- A theory for multiresolution signal decomposition: the wavelet representationIEEE Transactions on Pattern Analysis and Machine Intelligence, 1989
- An algorithm for the machine calculation of complex Fourier seriesMathematics of Computation, 1965