Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity
The imbalanced development between deep learning-based model design and motor imagery (MI) data acquisition raises concerns about the potential overfitting issue—models can identify training data well but fail to generalize test data. In this study, a Spatial Variation Generation (SVG) al...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Transactions on Neural Systems and Rehabilitation Engineering |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10248038/ |
| _version_ | 1827808889559580672 |
|---|---|
| author | Chengxuan Qin Rui Yang Mengjie Huang Weibo Liu Zidong Wang |
| author_facet | Chengxuan Qin Rui Yang Mengjie Huang Weibo Liu Zidong Wang |
| author_sort | Chengxuan Qin |
| collection | DOAJ |
| description | The imbalanced development between deep learning-based model design and motor imagery (MI) data acquisition raises concerns about the potential overfitting issue—models can identify training data well but fail to generalize test data. In this study, a Spatial Variation Generation (SVG) algorithm for MI data augmentation is proposed to alleviate the overfitting issue. In essence, SVG generates MI data using variations of electrode placement and brain spatial pattern, ultimately elevating the density of the raw sample vicinity. The proposed SVG prevents models from memorizing the training data by replacing the raw samples with the proper vicinal distribution. Moreover, SVG generates a uniform distribution and stabilizes the training process of models. In comparison studies involving five deep learning-based models across eight datasets, the proposed SVG algorithm exhibited a notable improvement of 0.021 in the area under the receiver operating characteristic curve (AUC). The improvement achieved by SVG outperforms other data augmentation algorithms. Further results from the ablation study verify the effectiveness of each component of SVG. Finally, the studies in the control group with varying numbers of samples show that the SVG algorithm consistently improves the AUC, with improvements ranging from approximately 0.02 to 0.15. |
| first_indexed | 2024-03-11T22:34:26Z |
| format | Article |
| id | doaj.art-b293a435b54047d288aab8836db0956b |
| institution | Directory Open Access Journal |
| issn | 1558-0210 |
| language | English |
| last_indexed | 2024-03-11T22:34:26Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Transactions on Neural Systems and Rehabilitation Engineering |
| spelling | doaj.art-b293a435b54047d288aab8836db0956b2023-09-22T23:00:05ZengIEEEIEEE Transactions on Neural Systems and Rehabilitation Engineering1558-02102023-01-01313675368610.1109/TNSRE.2023.331467910248038Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample VicinityChengxuan Qin0https://orcid.org/0009-0009-8463-3457Rui Yang1https://orcid.org/0000-0002-5634-5476Mengjie Huang2https://orcid.org/0000-0001-8163-8679Weibo Liu3https://orcid.org/0000-0002-8169-3261Zidong Wang4https://orcid.org/0000-0002-9576-7401School of Advanced Technology, Xi’an Jiaotong–Liverpool University, Suzhou, ChinaSchool of Advanced Technology, Xi’an Jiaotong–Liverpool University, Suzhou, ChinaDesign School, Xi’an Jiaotong–Liverpool University, Suzhou, ChinaDepartment of Computer Science, Brunel University London, Uxbridge, Middlesex, U.KDepartment of Computer Science, Brunel University London, Uxbridge, Middlesex, U.KThe imbalanced development between deep learning-based model design and motor imagery (MI) data acquisition raises concerns about the potential overfitting issue—models can identify training data well but fail to generalize test data. In this study, a Spatial Variation Generation (SVG) algorithm for MI data augmentation is proposed to alleviate the overfitting issue. In essence, SVG generates MI data using variations of electrode placement and brain spatial pattern, ultimately elevating the density of the raw sample vicinity. The proposed SVG prevents models from memorizing the training data by replacing the raw samples with the proper vicinal distribution. Moreover, SVG generates a uniform distribution and stabilizes the training process of models. In comparison studies involving five deep learning-based models across eight datasets, the proposed SVG algorithm exhibited a notable improvement of 0.021 in the area under the receiver operating characteristic curve (AUC). The improvement achieved by SVG outperforms other data augmentation algorithms. Further results from the ablation study verify the effectiveness of each component of SVG. Finally, the studies in the control group with varying numbers of samples show that the SVG algorithm consistently improves the AUC, with improvements ranging from approximately 0.02 to 0.15.https://ieeexplore.ieee.org/document/10248038/Brain–computer interfacesdata augmentationdeep learningelectroencephalogram |
| spellingShingle | Chengxuan Qin Rui Yang Mengjie Huang Weibo Liu Zidong Wang Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity IEEE Transactions on Neural Systems and Rehabilitation Engineering Brain–computer interfaces data augmentation deep learning electroencephalogram |
| title | Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity |
| title_full | Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity |
| title_fullStr | Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity |
| title_full_unstemmed | Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity |
| title_short | Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity |
| title_sort | spatial variation generation algorithm for motor imagery data augmentation increasing the density of sample vicinity |
| topic | Brain–computer interfaces data augmentation deep learning electroencephalogram |
| url | https://ieeexplore.ieee.org/document/10248038/ |
| work_keys_str_mv | AT chengxuanqin spatialvariationgenerationalgorithmformotorimagerydataaugmentationincreasingthedensityofsamplevicinity AT ruiyang spatialvariationgenerationalgorithmformotorimagerydataaugmentationincreasingthedensityofsamplevicinity AT mengjiehuang spatialvariationgenerationalgorithmformotorimagerydataaugmentationincreasingthedensityofsamplevicinity AT weiboliu spatialvariationgenerationalgorithmformotorimagerydataaugmentationincreasingthedensityofsamplevicinity AT zidongwang spatialvariationgenerationalgorithmformotorimagerydataaugmentationincreasingthedensityofsamplevicinity |