Machine learning based feature extraction from EEG signals for brain-computer interface
Over the recent years, Electroencephalography (EEG) signal analysis has been found is one of the most popular and powerful methods to study human’s physical activities. The EEG signal is a testing record which is utilized to reflect the electrical activity of brain cells in the cerebral cortex. By a...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
2018
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/74877 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | Over the recent years, Electroencephalography (EEG) signal analysis has been found is one of the most popular and powerful methods to study human’s physical activities. The EEG signal is a testing record which is utilized to reflect the electrical activity of brain cells in the cerebral cortex. By analyzing and transferring EEG signals to computer with BCI, brain activity can be translated and interpreted into instructions and commands for computer. As what we know, low spatial resolution and electrodes that catch signals overlapped with each other are disadvantages of EEG signals. Spatial filtering with dimensionality reduction is widely applied to alleviate overfitting problem and extract distinct features for Brain-Computer Interface (BCI). However, common spatial patterns (CSP) which is a widely used algorithm uses only a minimal number of spatial filters and ignores the rest. As a result, BCI performance is restricted. In this project, a novel feature weighting and regularization (FWR) method which uses all CSP features and its variants is applied with CSP to prevent BCI performance degradation. The experimental results demonstrate that the applied method FWR increases the classification accuracy comparing to existing CSP algorithm by working on BCI Competition III Dataset IVa and BCI Competition IV Dataset IIa. |
---|