Evaluation of Fast Sample Entropy Algorithms on FPGAs: From Performance to Energy Efficiency

Area of Science:

• Computational complexity analysis
• Biomedical signal processing
• Hardware acceleration

Background:

• Sample entropy (SampEn) is crucial for estimating time-series complexity but computationally intensive for large datasets.
• Existing SampEn algorithms face limitations in processing large volumes of data due to quadratic time complexity.
• Hardware acceleration offers a potential solution to overcome computational bottlenecks.

Purpose of the Study:

• To evaluate hardware architectures for accelerating sample entropy (SampEn) calculations.
• To assess the performance and energy efficiency of optimized SampEn algorithms on FPGAs.
• To compare FPGA-based SampEn implementations against CPU-based approaches.

Main Methods:

• Implementation of straightforward (SF), bucket-assist (BA), and lightweight (BS-LW, MS-LW) SampEn algorithms.
• Deployment of algorithms on embedded CPU, high-performance CPU, and Field-Programmable Gate Arrays (FPGAs).
• Evaluation using simulated data and real-world electrocardiogram (ECG) signals, analyzing execution time, resource usage, power, and energy consumption.

Main Results:

• FPGA implementations of SampEn algorithms demonstrate significant reductions in energy consumption (one to two orders of magnitude) compared to high-performance CPUs.
• Optimized SampEn algorithms, including BA, BS-LW, and MS-LW, were successfully implemented and evaluated on various hardware platforms.
• While not always significantly faster than high-performance CPUs in execution time, FPGAs offer superior power efficiency.

Conclusions:

• FPGA-based hardware acceleration is a viable and energy-efficient approach for complex time-series analysis using SampEn.
• Optimized SampEn algorithms deployed on FPGAs provide substantial power savings, making them suitable for large-scale data processing.
• This work highlights the potential of reconfigurable technologies for efficient computation of complexity metrics in applications like ECG analysis.