Feature selection of EEG data with neuro-statistical method

Author

Hossain, Md Zakir ; Kabir, Muammar M. ; Shahjahan, Md

Author_Institution

Dept. of Electr. & Electron. Eng., Khulna Univ. of Eng. & Technol., Khulna, Bangladesh

fYear

2014

fDate

13-15 Feb. 2014

Firstpage

1

Lastpage

6

Abstract

Feature selection (FS) of high dimensional electroencephalographic (EEG) data helps to identify and diagnose the brain conditions easily. Features can be selected with different ways where canonical correlation analysis (CCA) is one of them which are a statistical method. We employed neural network (NN) with CCA for salient features extraction of EEG data, called Neural Canonical Correlation Analysis (NCCA), which exhibits better result than individual CCA or NN. A NN classifier is used to test the classification of the selected features. The NN classifier shows remarkable result in terms of recognition rate.

Keywords

brain; electroencephalography; feature extraction; feature selection; medical signal processing; neural nets; neurophysiology; signal classification; statistical analysis; EEG data; NCCA; NN classifier; brain conditions; canonical correlation analysis; feature selection; features extraction; high-dimensional electroencephalographic data; neural canonical correlation analysis; neural network; neurostatistical method; patient diagnosis; recognition rate; Accuracy; Artificial neural networks; Brain models; Correlation; Electroencephalography; Neurons; Clustering; Electroencephalogram (EEG); Feature selection (FS); Neural Canonical Correlation Analysis (NCCA); Neural network (NN);

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical Information and Communication Technology (EICT), 2013 International Conference on

Conference_Location

Khulna

Print_ISBN

978-1-4799-2297-0

Type

conf

DOI

10.1109/EICT.2014.6777880

Filename

6777880