A new blind source separation approach based on dynamical similarity and its application on epileptic seizure prediction

Hamid Niknazar | Ali Motie-Nasrabadi | Mohammad Bagher Shamsollahi

---

URL :   http://research.shahed.ac.ir/WSR/WebPages/Report/PaperView.aspx?PaperID=158481
Date :  2021/02/17
Publish in :    Signal Processing
DOI :  https://doi.org/https://doi.org/10.1016/j.sigpro.2021.108045
Link :  http://dx.doi.org/https://doi.org/10.1016/j.sigpro.2021.108045
Keywords :Blind source separation, Seizure prediction, Nonlinear dynamics, Similarity measure, Meta heuristic search

Abstract :

---

Blind source separation is an important field of study in signal processing, in which the goal is to estimate source signals by having mixed observations. There are some conventional methods in this field that aim to estimate source signals by considering certain assumptions on sources. One of the most popular assumptions is the non-Gaussianity of sources which is the basis of many popular blind source separation methods. These methods may fail to estimate sources when the distribution of two or more sources is Gaussian. Hence, this study aims to introduce a new approach in blind source separation for nonlinear and chaotic signals by using a dynamical similarity measure and relaxing non-Gaussianity assumption. The proposed approach assumes there are dynamical stability in source signals and dynamical independence between them. The efficiency of the proposed approach is evaluated by synthetic simulation. Also, to evaluate the ability of the method in real-world applications and featuring its flexibility, the proposed approach is employed in epileptic seizure prediction by using EEG signals. The results show the potential and ability of the proposed method in nonlinear and chaotic signal processing.