Single-lead f-wave extraction using diffusion geometry

摘要: A novel single-lead f-wave extraction algorithm based on the modern diffusion geometry data analysis framework is proposed. The essentially an averaged beat subtraction algorithm, where ventricular activity template estimated by combining a newly designed metric, "diffusion distance," and non-local Euclidean median non-linear manifold setup. We coined DD-NLEM. Two simulation schemes are considered, new DD-NLEM outperforms traditional algorithms, including average subtraction, principal component analysis, adaptive singular value cancellation, in different evaluation metrics with statistical significance. clinical potential shown real Holter signal, we introduce score to evaluate performance of algorithm.