A Biologically Inspired Movement Recognition System with Spiking Neural Networks for Ambient Assisted Living
Athanasios Passias1, Karolos-Alexandros Tsakalos1, Ioannis Kansizoglou2
1Department of Electrical and Computer Engineering, Democritus University of Thrace, 67100 Xanthi, Greece.
Biomimetics (Basel, Switzerland)
|May 24, 2024
Summary
This study introduces energy-efficient spiking neural networks (SNNs) for ambient assisted living (AAL), achieving 83.4% accuracy in elderly movement recognition. This neuromorphic approach offers a practical alternative to traditional deep neural networks for eldercare applications.
Area of Science:
- Neuromorphic Engineering
- Artificial Intelligence
- Biomedical Engineering
Background:
- Increasing elderly populations necessitate advanced Ambient Assisted Living (AAL) solutions.
- Traditional Deep Neural Networks (DNNs) present energy and computational challenges for AAL.
- Spiking Neural Networks (SNNs) offer a more energy-efficient, biologically inspired alternative.
Purpose of the Study:
- To develop an energy-efficient SNN-based system for AAL applications.
- To investigate the efficacy of asynchronous cellular automaton-based neurons (ACANs) for SNNs.
- To enhance SNN learning efficiency using the remote supervised method (ReSuMe) for movement recognition.
Main Methods:
- Implementation of asynchronous cellular automaton-based neurons (ACANs) for hardware efficiency.
- Application of the remote supervised method (ReSuMe) to improve spike train learning in SNNs.
- Utilizing motion capture data for movement recognition in an elderly population.
Main Results:
- Achieved high classification accuracy of 83.4% in elderly movement activity recognition.
- Demonstrated superior computational efficiency of SNNs compared to conventional DNNs.
- Validated the potential for real-time, energy-efficient processing in AAL environments.
Conclusions:
- The proposed SNN approach using ACANs and ReSuMe is effective for AAL.
- This neuromorphic solution offers significant advantages in energy efficiency and processing for eldercare.
- The findings support the practical application of neuromorphic computing in AAL systems.
