The implementation of EEG transfer learning method using integrated selection for motor imagery signal
Brain-computer interface (BCI) is a system that can translate, manage, and recognize human brain activity. One of the devices used in the BCI system is Electroencephalogram (EEG). The brain signals produced by the EEG are diverse. One of them is the motor imagery signal. Motor imagery signal is used...
| Main Authors: | , , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Published: |
2021
|
| Subjects: |
| _version_ | 1796866086715523072 |
|---|---|
| author | Ramadhani, A. Fauzi, H. Wijayanto, I. Rizal, A. Shapiai, M. I. |
| author_facet | Ramadhani, A. Fauzi, H. Wijayanto, I. Rizal, A. Shapiai, M. I. |
| author_sort | Ramadhani, A. |
| collection | ePrints |
| description | Brain-computer interface (BCI) is a system that can translate, manage, and recognize human brain activity. One of the devices used in the BCI system is Electroencephalogram (EEG). The brain signals produced by the EEG are diverse. One of them is the motor imagery signal. Motor imagery signal is used to translate the EEG signal into a specific movement. The performance of motor imagery signal classification is influenced by the number of training and testing data used. In most cases, the training data consists of a higher number of trials than the testing data. However, more trials cause higher subject variation. Previously study mentioned that this problem can be overcome by using transfer learning methods, which aimed at simplifying the training model. In this study, transfer learning in BCI is implemented using the integrated selection (IS) method, which simplifies the training model. Furthermore, IS is optimizing the data by removing the irrelevant channels of the EEG signals. Integrated selection uses the CUR matrix decomposition algorithm. The method split the data into two components, namely identity and historical data, represented by the C and UR matrix, respectively. The characteristic of the data from IS then calculated using three feature extraction methods. They are Fast Fourier Transform (FFT), Hjorth Descriptor, and Common Spatial Pattern (CSP). The features are then classified using the k-Nearest Neighbor (K-NN) method. The use of IS in the BCI system increases the accuracy of more than 6% and six-times faster processing time. In general, the integrated selection method is able to improve the performance of the BCI system. |
| first_indexed | 2024-03-05T21:06:55Z |
| format | Conference or Workshop Item |
| id | utm.eprints-95760 |
| institution | Universiti Teknologi Malaysia - ePrints |
| last_indexed | 2024-03-05T21:06:55Z |
| publishDate | 2021 |
| record_format | dspace |
| spelling | utm.eprints-957602022-05-31T13:18:49Z http://eprints.utm.my/95760/ The implementation of EEG transfer learning method using integrated selection for motor imagery signal Ramadhani, A. Fauzi, H. Wijayanto, I. Rizal, A. Shapiai, M. I. T Technology (General) Brain-computer interface (BCI) is a system that can translate, manage, and recognize human brain activity. One of the devices used in the BCI system is Electroencephalogram (EEG). The brain signals produced by the EEG are diverse. One of them is the motor imagery signal. Motor imagery signal is used to translate the EEG signal into a specific movement. The performance of motor imagery signal classification is influenced by the number of training and testing data used. In most cases, the training data consists of a higher number of trials than the testing data. However, more trials cause higher subject variation. Previously study mentioned that this problem can be overcome by using transfer learning methods, which aimed at simplifying the training model. In this study, transfer learning in BCI is implemented using the integrated selection (IS) method, which simplifies the training model. Furthermore, IS is optimizing the data by removing the irrelevant channels of the EEG signals. Integrated selection uses the CUR matrix decomposition algorithm. The method split the data into two components, namely identity and historical data, represented by the C and UR matrix, respectively. The characteristic of the data from IS then calculated using three feature extraction methods. They are Fast Fourier Transform (FFT), Hjorth Descriptor, and Common Spatial Pattern (CSP). The features are then classified using the k-Nearest Neighbor (K-NN) method. The use of IS in the BCI system increases the accuracy of more than 6% and six-times faster processing time. In general, the integrated selection method is able to improve the performance of the BCI system. 2021 Conference or Workshop Item PeerReviewed Ramadhani, A. and Fauzi, H. and Wijayanto, I. and Rizal, A. and Shapiai, M. I. (2021) The implementation of EEG transfer learning method using integrated selection for motor imagery signal. In: 1st International Conference on Electronics, Biomedical Engineering, and Health Informatics, ICEBEHI 2020, 8 October 2020 - 9 October 2020, Virtual, Online. http://dx.doi.org/10.1007/978-981-33-6926-9_39 |
| spellingShingle | T Technology (General) Ramadhani, A. Fauzi, H. Wijayanto, I. Rizal, A. Shapiai, M. I. The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title | The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title_full | The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title_fullStr | The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title_full_unstemmed | The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title_short | The implementation of EEG transfer learning method using integrated selection for motor imagery signal |
| title_sort | implementation of eeg transfer learning method using integrated selection for motor imagery signal |
| topic | T Technology (General) |
| work_keys_str_mv | AT ramadhania theimplementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT fauzih theimplementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT wijayantoi theimplementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT rizala theimplementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT shapiaimi theimplementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT ramadhania implementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT fauzih implementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT wijayantoi implementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT rizala implementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal AT shapiaimi implementationofeegtransferlearningmethodusingintegratedselectionformotorimagerysignal |