An ultra low energy biomedical signal processing system operating at near-threshold

Area of Science:

• Biomedical Engineering
• Digital Signal Processing
• Low-Power Electronics

Background:

• Ambulatory monitoring of biomedical signals requires low-power wireless sensor nodes (WSNs) for extended battery life and reduced size.
• Existing systems often struggle to balance performance with energy efficiency for complex signal processing tasks.

Purpose of the Study:

• To develop a voltage-scalable digital signal processing system for WSNs tailored for energy-aware ambulatory monitoring.
• To create a flexible processing platform capable of adapting power and performance to specific application needs.

Main Methods:

• Designed an event-driven architecture incorporating system partitioning, duty cycling, SIMD instructions, power gating, voltage scaling, multiple clock/voltage domains, and clock gating.
• Implemented a case study using a continuous wavelet transform (CWT) for ElectroCardioGram (ECG) heart-beat detection.

Main Results:

• The proposed platform demonstrated effective energy scaling for varying application complexities.
• Achieved the lowest reported energy per sample for ECG heart-beat detection in ambulatory monitoring.
• Maintained the sensitivity and positive predictivity of the CWT-based heart-beat detection algorithm.

Conclusions:

• The developed voltage-scalable DSP system offers a significant advancement in energy efficiency for WSNs in biomedical applications.
• This platform provides a viable solution for long-term, low-power ambulatory monitoring of physiological signals.
• The system's adaptability and performance pave the way for more sophisticated WSN-based healthcare solutions.