Acceleration of Bucket-Assisted Fast Sample Entropy for Biomedical Signal Analysis

Sample entropy (SampEn) is widely used to assess the complexity of physiological time-series signals. However, it is a computationally intensive algorithm with O(N-2) time complexity. Although algorithmic optimizations, such as bucket-assisted SampEn, have been proposed to eliminate unnecessary computations, the time demand restricts their use in real-time applications with long-term inputs. To address the time and space complexity issue in SampEn, we optimize bucket-assisted SampEn by dynamic memory allocation to avoid space complexity and accelerate the optimized bucket-assisted SampEn on field-programmable gate arrays (FPGAs). Our method accelerates bucket-assisted SampEn through efficient random storage and data access on FPGA. Furthermore, we introduce a scheduling strategy to handle unbalanced loads for time-intensive inter- and intrasimilarity comparisons. We validate our approach on multisource biomedical signals and demonstrate its effectiveness by achieving more than two orders of magnitude faster than straightforward (SF) defined SampEn computation. Our work provides a practical and effective approach for measuring time-series complexity using bucket-assisted SampEn on FPGA, with the potential for real-time applications with long-term inputs.

Automated summary

This study optimizes the bucket-assisted SampEn algorithm to address its time and space complexity issue, and accelerates it on FPGA through efficient random storage and data access. A scheduling strategy is introduced to handle unbalanced loads. Experimental results show that our approach is effective and practical for measuring time-series complexity.