Study of non-invasive cognitive tasks and feature extraction techniques for brain-computer interface (BCI) applications
A brain-computer interface (BCI) provides an important alternative for disabled people that enables the non-muscular communication pathway among individual thoughts and different assistive appliances. A BCI technology essentially consists of data acquisition, pre-processing, feature extraction, clas...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/34407/1/Study%20of%20non-invasive%20cognitive%20tasks%20and%20feature.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A brain-computer interface (BCI) provides an important alternative for disabled people that enables the non-muscular communication pathway among individual thoughts and different assistive appliances. A BCI technology essentially consists of data acquisition, pre-processing, feature extraction, classification and device command. Indeed, despite the valuable and promising achievements already obtained in every component of BCI, the BCI field is still a relatively young research field and there is still much to do in order to make BCI become a mature technology. To mitigate the impediments concerning BCI, the study of cognitive task together with the EEG feature and classification framework have been investigated. There are four distinct experiments have been conducted to determine the optimum solution to those specific issues. In the first experiment, three cognitive tasks namely quick math solving, relaxed and playing games have been investigated. The features have been extracted using power spectral density (PSD), logenergy entropy, and spectral centroid and the extracted feature has been classified through the support vector machine (SVM), K-nearest neighbor (K-NN), and linear discriminant analysis (LDA). In this experiment, the best classification accuracy for single channel and five channel datasets were 86% and 91.66% respectively that have been obtained by the PSD-SVM approach. The wink based facial expressions namely left wink, right wink and no wink have been studied through fast Fourier transform (FFT) and sample range feature and then the extracted features have been classified using SVM, K-NN, and LDA. The best accuracy (98.6%) has been achieved by the sample range-SVM based approach. The eye blinking based facial expression has been investigated following the same methodology as the study of wink based facial expression. Moreover, the peak detection approach has also been employed to compute the number of blinks. The optimum accuracy of 99% has been achieved using the peak detection approach. Additionally, twoclass motor imagery hand movement has been classified using SVM, K-NN, and LDA where the feature has been extracted through PSD, spectral centroid and continuous wavelet transform (CWT). The optimum 74.7% accuracy has been achieved by the PSDSVM approach. Finally, two device command prototypes have been designed to translate the classifier output. One prototype can translate four types of cognitive tasks in terms of 5 watts four different colored bulbs, whereas, another prototype may able to control DC motor utilizing cognitive tasks. This study has delineated the implementation of every BCI component to facilitate the application of brainwave assisted assistive appliances. Finally, this thesis comes to the end by drawing the future direction regarding the current issues of BCI technology and these directions may significantly enhance usability for the implementation of commercial applications not only for the disabled but also for a significant number of healthy users. |
|---|