Efficient implementation of LMS adaptive filter-based FECG extraction on an FPGA

Area of Science:

• Biomedical Engineering
• Digital Signal Processing
• Hardware Implementation

Background:

• Accurate fetal heart rate (FHR) monitoring is crucial for prenatal care.
• Existing methods for fetal electrocardiogram (FECG) extraction face challenges in noise reduction and computational efficiency.
• Field Programmable Gate Arrays (FPGAs) offer a flexible platform for developing real-time biomedical signal processing systems.

Purpose of the Study:

• To develop and implement an FPGA-based system for fetal heart rate (FHR) monitoring.
• To enhance the accuracy and precision of FECG extraction and FHR detection.
• To propose and evaluate two distinct architectures for the Least Mean Squares Adaptive Filter (LMS-AF) for FECG extraction.

Main Methods:

• Implementation of a preprocessing unit for noise removal.
• Development of a foetal electrocardiogram (FECG) extraction unit utilizing a Least Mean Squares Adaptive Filter (LMS-AF).
• Design of two LMS-AF architectures (series and parallel) on FPGA, including a floating-point unit for improved arithmetic operations.

Main Results:

• The system achieved high sensitivity (95.74-100%) and specificity (100%) in detecting R peaks from extracted FECG.
• The parallel LMS-AF architecture demonstrated up to an 85.88% reduction in convergence time.
• The series LMS-AF architecture showed a 27.41% reduction in hardware resource utilization (flip-flops).
• An overall accuracy increase of 2-7.51% was observed compared to previous works.

Conclusions:

• The proposed FPGA-based system provides an effective and efficient solution for fetal heart rate monitoring.
• The parallel architecture optimizes convergence time and power consumption, while the series architecture minimizes hardware resource usage.
• This work contributes to advancing non-invasive FECG extraction techniques with improved performance metrics.