FPGA-Based Real-Time Quality Analysis Framework for Photoplethysmogram Signals

This study presents the hardware implementation and performance assessment of a unified System-on-Chip (SoC) framework for modular digital signal processing (DSP) on an FPGA. Ten high-level synthesis (HLS)-based IP cores are included in the framework, which is aimed at biomedical applications, specifically photoplethysmogram (PPG) signal analysis, on a Xilinx Zynq-7000 SoC. The developed IP library comprises modules for mean removal, first-order derivative filtering, zerophase IIR filtering, amplitude normalization, and feature extraction. Evaluation was performed on 1,340 PPG segments (670 clean and 670 noisy) from the BIDMC, SLP, PulseDB and Wrist PPG databases using four processing pipelines: time-domain, derivative-based, FFT-based, and STFT-based. A menu-driven UART-based SDK interface enables performance benchmarking, cycle-accurate monitoring, and seamless hardware-software interaction. To guaranty improved performance, each IP core is developed analytically, implemented with Vivado HLS, and verified against MATLAB reference output. Using derivativebased skewness characteristics, a thorough investigation of signal quality analysis provides up to 99% accuracy in classifying clean and noisy signals. The synthesis result at 100 MHz shows that the compute-intensive spectral modules offer detailed time-frequency characterization, with a total SoC power of less than 2 W and each IP less than 160 mW. The light modules (mean removal and derivative filtering) have latencies as low as 100–125 cycles. Design optimizations such as memory partitioning, function inlining, and loop pipelining further improve throughput and resource efficiency. The suggested SoC framework provides a hardware base that is real-time, reusable, and power-efficient for biomedical signal analytics. Bridges the gap between algorithm creation and embedded implementation. The framework enables physiological monitoring, edge-level health analytics, and robust PPG quality assessment on the device.