The extraction of motion-onset VEP BCI features based on deep learning and compressed sensing

详细信息    查看全文

• 作者：Teng Ma^a ; Hui Li^a ; Hao Yang^a ; Xulin Lv^a ; Peiyang Li^a ; Tiejun Liu^a ; ^b ; Dezhong Yao^a ; ^b ; ^{dyao@uestc.edu.cn} ; Peng Xu^a ; ^b ; ^{xupeng@uestc.edu.cn}
• 关键词：Motion-onset VEP ; Brain computer interface ; Compressed sensing ; Deep learning ; Multi-modality feature
• 刊名：Journal of Neuroscience Methods
• 出版年：2017
• 出版时间：1 January 2017
• 年：2017
• 卷：275
• 期：Complete
• 页码：80-92
• 全文大小：4178 K
• 卷排序：275

文摘

Deep learning is combined with compressed sensing to mine discriminative mVEP information. The deep learning and compressed sensing approach can generate multi-modal features. The proposed multi-modal feature can effectively improve the BCI performance with approximately 3.5% accuracy incensement. The deep learning and compressed sensing approach is more effective for subjects with relatively poor performance.