Advanced techniques for classification of multi-channel EEG signals for brain computer interface /
Electroencephalogram (EEG) signal based research is ongoing for the development of simple, user friendly, robust, efficient brain computer interfacing (BCI) devices/systems. Motor imagery related EEG signal classification is one of the main challenges in designing of a BCI system. An advanced and si...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur :
Kulliyyah of Engineering, International Islamic University Malaysia,
2015
|
| Subjects: | |
| Online Access: | Click here to view 1st 24 pages of the thesis. Members can view fulltext at the specified PCs in the library. |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Electroencephalogram (EEG) signal based research is ongoing for the development of simple, user friendly, robust, efficient brain computer interfacing (BCI) devices/systems. Motor imagery related EEG signal classification is one of the main challenges in designing of a BCI system. An advanced and simple classification technique for motor imagery related BCI system has been developed. Fisher Linear Discriminant Analysis (FLDA) has a very low computational requirement, which makes it suitable for BCI system. Motor imagery based EEG dataset, collected by the world renowned BCI Group from Graz University of Technology, Austria, has been used. Initially the signal is extracted into features. The power spectral density technique has been used to extract the non-linear features over some frequency components in motor imagery based EEG signals. In training phase FLDA, Mahalanobis Discriminant Analysis (MDA), Quadratic Discriminant Analysis (QDA), Cauchy and Gaussian Radial Basis (GRB) classification techniques have been used for designing a motor imagery based BCI system. Then the optimized classifier MDA has been chosen. But in evaluation phase, FLDA performs better than MDA. This classification technique separates the extracted signals into possible classes by taking the means and variances between two classes. Then percentage of accuracy has been measured to detect the motor imagery movement. In addition, the probabilistic accuracy has been measured by using Cohen's kappa. It obtains more than 98% of accuracy and around 95% kappa in average in training phase using MDA. In evaluation state, they are 46% and 39% accuracies using FLDA and MDA respectively. For contrastive justification, some other classification techniques have also been used to compare the obtained results. Result depicts that the MDA classifier could be a preferable classification technique for both training and evaluation for detecting different motor imagery related brain states. |
|---|---|
| Physical Description: | xv, 95 leaves : ill ; 30cm. |
| Bibliography: | Includes bibliographical references (leave 85-93). |