Motor Imagery EEG Classification with Biclustering Based Fuzzy Inference
-
Published:2020-07-01
Issue:7
Volume:10
Page:1486-1493
-
ISSN:2156-7018
-
Container-title:Journal of Medical Imaging and Health Informatics
-
language:en
-
Short-container-title:j med imaging hlth inform
Affiliation:
1. School of Computer Science, Northwestern Polytechnical University, Xi’an 710072, Shaanxi, People’s Republic of China
Abstract
The rehabilitation of armless or footless patients is of great importance. One choice for such issue is using the electroencephalograph (EEG) brain computer interface to help the patients communicate with outside. Classifying the EEG signals generated from mental activity is one of
the most important technologies. However, existing classification methods often suffer the overfitting problem caused by the small training data sets while big dimensionality of feature space. Fuzzy inference can imitate the human judgement, effectively dealing with uncertainty and small-sample
learning problems. Besides, biclustering has shown excellent performance in constructing rule base. This paper proposes a novel biclustering based fuzzy inference method for EEG classification. It can be divided into five steps. The first step is generating features with common spatial pattern.
The second step is searching local coherent patterns with column nearly constant biclustering. The third step is to transform the patterns to if-then rules with column averaging and majority voting strategy. Subsequent step is to employ Mamdani fuzzy inference to map the input feature vector
into decimals. Finally, particle swarm optimization is utilized to generate optimal threshold for linear classification. Experiments on several commonly used data sets show that the proposed method has advantages over competitors in terms of classification accuracy.
Publisher
American Scientific Publishers
Subject
Health Informatics,Radiology, Nuclear Medicine and imaging
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献