Single trial classification of EEG in predicting intention and direction of wrist movement : translation toward development of four-class brain computer interface system based on a single limb

Syam, Syahrull Hi Fi and Lakany, Heba and Conway, BA; (2016) Single trial classification of EEG in predicting intention and direction of wrist movement : translation toward development of four-class brain computer interface system based on a single limb. In: 8th International Conference on Advanced Cognitive Technologies and Applications. Curran Associates, Inc., ITA, pp. 90-95. ISBN 9781510821156

[thumbnail of Syam-etal-COGNITIVE-2016-Single-trial-classification-of-EEG-in-predicting-intention]
Preview
 Text. Filename: Syam_etal_COGNITIVE_2016_Single_trial_classification_of_EEG_in_predicting_intention.pdf
Final Published Version
Download (420kB)| Preview

Abstract

Brain - computer interfaces (BCI) are paradigms that offer an alternative communication channel between neural activity gene rated in the brain and the users’ external environment. The aim of this paper is to investigate the feasibility of designing and developing a multiclass BCI system based on a single limb movement due to the factor, high dimensional control channels would expand the capacity of BCI application (multidimensional control of neuroprosthesis). This paper also proposes a method to identify the optimal frequency band and recording channel to achieve the best classification result . Twenty eight surface electroencephalography ( EEG ) electrodes are used to record brain activity from eleven subjects whilst imagining and performing right wrist burst point - to - point movement towards multiple directions using a high density montage with 10 - 10 electrode placement locations focusing on motor cortex areas. Two types of spatial filters namely Common average reference (CAR) and Laplacian (LAP) filter have been implemented and results are compared to enhance the EEG signal. Features are extracted from the filtered signals using event related spectral perturbation ( ERSP ) and power spectrum. Feature vectors are classified by k - nearest neighbour ( k - NN) and quadratic discriminant analysis (QDA) classifiers. The results indicate that the majority of the optimum classification results are obtained from features extracted from contralateral electrodes in the gamma band. Based on a single trial, the average of the classification accuracy using LAP filter and k - NN classifier across the subjects in predicting intention and direction of movement is 68% and 62% for motor imagery and motor performance respectively; which is significantly higher than chance. The classification result from the majority of subjects shows that, it is possible and achievable to develop multiclass BCI systems based on a single limb.

CORE (COnnecting REpositories)