Intermittent Neuromorphic Wearable Systems.

Area of Science:

• Biomedical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Wearable medical devices are crucial for chronic disease management, offering benefits like early detection and reduced healthcare costs.
• Integrating Artificial Intelligence (AI) enhances wearable capabilities but increases power demands, limiting device lifespan.
• High energy consumption of AI in wearables necessitates innovative power-saving solutions.

Purpose of the Study:

• To present a novel framework for reducing energy consumption in embedded AI for wearable devices.
• To address the power limitations of AI-driven wearables through advanced computing techniques.
• To demonstrate the framework's effectiveness in a practical application.

Main Methods:

• A novel framework combining neuromorphic computing and intermittent computing was developed.
• The framework was applied to an Electromyography (EMG) application for hand gesture classification.
• The Ninapro DB2 dataset was utilized for validating the proof-of-concept.

Main Results:

• The proposed framework successfully reduced embedded AI energy consumption.
• The framework demonstrated state-of-the-art energy efficiency for wearable AI.
• The application to EMG-based gesture recognition validated the framework's practical utility.

Conclusions:

• The novel framework effectively enhances energy efficiency in wearable AI devices.
• Combining neuromorphic and intermittent computing offers a promising solution for power-hungry AI wearables.
• This approach supports the development of sustainable and long-lasting AI-powered wearable medical technology.