| Title:
|
A Deep Learning Approach for Classifying Four-Class Motor Imagery Using EEG Signals
|
| Author(s):
|
1
|
| Subject:
|
Brain–Computer Interfaces (BCIs)
|
|
|
|
Presenter Author:
|
|
| Name:
|
AmirAbbas Amini
|
|
|
|
Author:
|
|
| Name 1:
|
AmirAbbas Amini - Corresponding - Presenter
|
| Affiliation 1:
|
School of Advanced Technologies in Medicine, Medical Image and Signal Processing Research Center, Isfahan University of Medical Sciences, Isfahan, Iran
|
|
|
| Name 2:
|
|
| Affiliation 2:
|
|
|
|
| Name 3:
|
|
| Affiliation 3:
|
|
|
|
| Name 4:
|
|
| Affiliation 4:
|
|
|
|
| Name 5:
|
|
| Affiliation 5:
|
|
|
|
| Name 6:
|
|
| Affiliation 6:
|
|
|
|
| Name 7:
|
|
| Affiliation 7:
|
|
|
|
| Name 8:
|
|
| Affiliation 8:
|
|
|
|
| Name 9:
|
|
| Affiliation 9:
|
|
|
|
| Name 10:
|
|
| Affiliation 10:
|
|
|
|
Abstract
|
Original Research
|
|
introduction :
|
Brain-computer interfaces based on motor imagery offer valuable assistive solutions for individuals with motor disabilities. These systems allow control of external devices such as wheelchairs or robotic prosthetics through imagined movement alone, without physical muscle activity. Accurate classification of distinct motor imagery states remains a critical challenge in advancing practical BCI systems. This study presents an automated and efficient method to discriminate four motor imagery classes, including right hand, left hand, feet, and tongue, using electroencephalography signals.
|
|
Methods:
|
This study employed the publicly available BCI Competition IV dataset 2a. Raw EEG signals underwent preprocessing with a 4 to 40 Hz bandpass filter, and ocular artifacts were removed. Time-frequency features were extracted using the continuous wavelet transform from segmented signal intervals. A three-layer convolutional neural network was designed, where each layer consisted of convolution, max pooling, and ReLU activation operations. A fully connected layer followed by a softmax output layer completed the four-class classifier. For each subject, data were divided into 70% for training and 30% for testing. The Adam optimizer was used during model training.
|
|
Results:
|
The proposed model attained an average classification accuracy of 81%. The highest performance was observed for the right hand motor imagery class, with accuracy of 85%, sensitivity of 83%, and specificity of 91%. The left hand class achieved accuracy of 82%, sensitivity of 80%, and specificity of 89%. For the feet class, accuracy was 79%, sensitivity 77%, and specificity 88%. The tongue class yielded accuracy of 78%, sensitivity 76%, and specificity 87%. The overall F1 score of the system was 0.80.
|
|
Conclusion:
|
The results indicate that a convolutional neural network based deep learning approach can effectively classify motor imagery states for brain-computer interface applications. This method holds strong potential for real-world deployment in assistive technologies supporting individuals with motor impairments.
|
| Attachment:
|
|
| Keywords:
|
Brain-computer interface, motor imagery, electroencephalography, deep learning, convolutional neural network
|
| Key message:
|
|
| Do you want your abstract to be indexed, if this is selected as one of the top 100
articles?
|
