A subject-dependent brain-computer interface based on motor imagery mental strategy.
The dataset consists of EEG data from 9 subjects. The cue-based BCI paradigm
consisted of four different motor imagery tasks, namely the imagination of
movement of the left hand (class 1), right hand (class 2), both feet (class 3),
and tongue (class 4). Two sessions on different days were recorded for each
subject. Each session is comprised of 6 runs separated by short breaks. One
run consists of 48 trials (12 for each of the four possible classes), yielding a
total of 288 trials per session. 22 EEG channels, and 3 EOG channels
We apply 3 techniques for preprocessing the signals:
Band Pass Filter: Bandpass filtering is performed using a sixth order
Butterworth bandpass filter with low cut of 8 Hz and high cut of 30 Hz. This
choice is because of the fact that motor imagery features generally happen
in the alpha and beta bands of EEG.
Common Average Referencing (CAR) Spatial Filtering: enhances the local
activity at electrode I by subtracting the average over all electrodes.
Normalization: Each channel is used its own mean and standard deviation.
.png?raw=true)
.png?raw=true)
The main idea is to use a linear transform to project the multichannel EEG
data into low-dimensional spatial subspace with a projection matrix, of
which each row consists of weights for channels.
1- We used Build in CSP from MNE library to extract features
2- We used Wavelet to extract features
1- Random Forest
2- SVM ((decision_function_shape='ovo'), ( kernel='rbf',gamma=0.5,C=0.1), (kernel='poly',degree=3,C=1))
3- KNN (n_neighbors=4)
4- Logistic Regression (multi_class='ovr', solver='liblinear')
In CSP, The best accuracy is Logistic Regression with 86%
In wavelet, The best accuracy is Logistic Regression with 76%
.png?raw=true)
.png?raw=true)
.png?raw=true)
.png?raw=true)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent