Robust Motor Imagery Classification Using Sparse Representations and Grouping Structures
The classification of Motor Imagery (MI) tasks constitutes one of the most challenging problems in Brain Computer Interfaces (BCI) mostly due to the varying conditions of its operation. These conditions may vary with respect to the number of electrodes, the time and effort that can be invested by th...
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
IEEE
2020
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/60fbede0a5354170938cd0a139579945 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | The classification of Motor Imagery (MI) tasks constitutes one of the most challenging problems in Brain Computer Interfaces (BCI) mostly due to the varying conditions of its operation. These conditions may vary with respect to the number of electrodes, the time and effort that can be invested by the user for training/calibrating the system prior to its use, as well as the duration or even the type of the imaginary task that is most convenient for the user. Hence, it is desirable to design classification schemes that are not only accurate in terms of the classification output but also robust to changes in the operational conditions. Towards this goal, we propose a new sparse representation classification scheme that extends current sparse representation schemes by exploiting the group sparsity of relevant features. Based on this scheme each test signal is represented as a linear combination of train trials that are further constrained to belong in the same MI class. Our expectation is that this constrained linear combination exploiting the grouping structure of the training data will lead to representations that are more robust to varying operational conditions. Moreover, in order to avoid overfitting and provide a model with good generalization abilities we adopt the bayesian framework and, in particular, the Variational Bayesian Framework since we use a specific approximate posterior to exploit the grouping structure of the data. We have evaluated the proposed algorithm on two MI datasets using electroencephalograms (EEG) that allowed us to simulate different operational conditions like the number of available channels, the number of training trials, the type of MI tasks, as well as the duration of each trial. Results have shown that the proposed method presents state-of-the-art performance against well known classification methods in MI BCI literature. |
|---|