Real-time non-uniform EEG sampling. (September 2021)

• Record Type: Journal Article
• Authors: Alipour, Mahmoud; Hashemi Gholpayeghani, Seyed Mohammad Reza
• Is Part Of: Biomedical signal processing and control. Volume 70(2021)
• Publication Date: 2021-09
• Subjects: Non-uniform sampling -- Signal compression -- EEG -- Information theory -- Lempel-Ziv complexity -- Non-linear features; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102961
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: We improved the performance of the Lempel-Ziv algorithm by introducing a dynamic threshold level determination method. We introduced a new EEG non-uniform sampling and compression method, which can sample EEG in real-time. Our method can universally sample any EEG data without prior knowledge. The proposed method by increasing EEG information richness can enhance non-linear features distinguishability. Abstract: Lempel-Ziv algorithm can measure the complexity of a non-linear system. By improving the performance of this algorithm, we introduce a new non-uniform sampling method of EEG. Despite its low computational load for EEG compression, our method can sample it in real-time. The proposed method is a nonparametric technique that can compress EEG with a 1.97 compress ratio without prior knowledge, training, or preconditions. We evaluate the efficiency of our method in signal compression with common performance evaluation methods. By presenting another approach for evaluating the proposed method, we extract four non-linear features from EEG of sleep stages before and after applying our sampling method. The proposed method can increase non-linear feature distinguishability such as fractal dimension more than seven times by ignoring redundant samples. This study investigates the effect of redundant samples on EEG information richness. We demonstrate that our method by ignoring EEG useless samples can increase EEG information richness.