Low complexity underdetermined blind source separation system architecture for emerging remote healthcare...

Area of Science:

• Biomedical Signal Processing
• Algorithm Design
• VLSI Architecture

Background:

• Remote healthcare applications require efficient biomedical signal processing.
• Underdetermined Blind Source Separation (UBSS) is crucial but computationally intensive.
• Resource constraints (area, power) limit existing UBSS algorithms in portable devices.

Purpose of the Study:

• To propose a low-complexity architecture for UBSS algorithms.
• To optimize the computationally intensive N-point Discrete Hilbert Transform (DHT) module.
• To enable UBSS in power- and area-constrained remote healthcare applications.

Main Methods:

• Developed a novel low-complexity architecture for UBSS.
• Identified and redesigned the N-point Discrete Hilbert Transform (DHT) as the key bottleneck.
• Implemented and compared the proposed DHT architecture against state-of-the-art methods.

Main Results:

• The proposed DHT architecture achieves significant reductions in on-chip area (up to 50.28%) and power consumption (up to 53.25%) for N=32, 64, 128.
• The new DHT architecture supports N=2m points, offering greater flexibility than the state-of-the-art N=4m point design.
• The overall UBSS architecture is optimized for resource-constrained environments.

Conclusions:

• The proposed low-complexity UBSS architecture, particularly the optimized DHT module, is highly suitable for remote healthcare.
• Significant power and area savings make this design practical for battery-operated biomedical devices.
• This advancement facilitates more sophisticated biomedical signal processing in remote monitoring scenarios.