A neuromorphic multi-scale approach for real-time heart rate and state detection

Area of Science:

• Biomedical Engineering
• Neuromorphic Computing
• Wearable Technology

Background:

• Wearable systems for continuous biosignal monitoring face challenges in size and energy consumption for long-term trend detection.
• Existing smartwatches offer high-precision heart rate tracking but struggle with always-on, multi-scale analysis.
• Neuromorphic technologies offer a potential solution due to their ultra-low power consumption.

Purpose of the Study:

• To propose and validate an energy-efficient biosignal processing architecture for wearable health monitoring.
• To leverage mixed-signal neuromorphic technology for continuous, multi-scale analysis of physiological data.
• To enable the development of stand-alone wearable devices for long-term health assessment.

Main Methods:

• Developed a biosignal processing architecture integrating multimodal sensory inputs.
• Utilized principles of neural computation for processing physiological data.
• Validated the architecture on a mixed-signal neuromorphic processor, assessing its robustness with analog circuit variability.
• Demonstrated multi-scale signal processing, including instantaneous heart rate and long-term physiological states.

Main Results:

• The proposed architecture reliably detects trends in heart rate and physiological states.
• The system demonstrated robust operation on a neuromorphic processor despite analog circuit variability.
• Successfully processed multi-scale signals, identifying monotonic changes indicative of pathological conditions like agitation.
• Achieved effective detection of long-term physiological state changes over extended periods.

Conclusions:

• The developed neuromorphic architecture offers an energy-efficient solution for continuous wearable health monitoring.
• This approach enables a new generation of stand-alone wearable devices suitable for long-term health assessment with minimal maintenance.
• The system's ability to process multi-scale signals and detect pathological trends paves the way for advanced health monitoring applications.