Identification of motor imagery tasks using power-based connectivity descriptors from EEG signals
In recent years, functional connectivity has been studied through electroencephalography signals to analyze the patterns generated by the electrical conductions of the brain. In BCI systems, the paradigm of motor imagery has been used to generate patterns to identify the user’s intention. However, t...
Guardado en:
| Autor principal: | |
|---|---|
| Otros Autores: | |
| Formato: | Trabajo de grado (Pregrado y/o Especialización) |
| Lenguaje: | eng |
| Publicado: |
Universidad Antonio Nariño
2022
|
| Materias: | |
| Acceso en línea: | http://repositorio.uan.edu.co/handle/123456789/6578 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | In recent years, functional connectivity has been studied through electroencephalography signals to analyze the patterns generated by the electrical conductions of the brain. In BCI systems, the paradigm of motor imagery has been used to generate patterns to identify the user’s intention. However, the study of techniques that allow the correct identification and classification of such intention is still a challenge due to the low performance of algorithms for rehabilitation engineering applications. This study addresses the problem of binary identification of left and right-hand opening and closing motor imagery tasks. A method called Power-Based Connectivity (PBC) is proposed that correlates two reference channels in the central cortex (C3 and C4) with other channels located in the central area of the brain. The methods were evaluated using an EEG dataset of six subjects with no previous experience in BCI systems built at the Antonio Narino University. The method was compared ˜ with a standard feature extraction method based on Power Spectral Density (PSD). It was used for evaluation accuracy and cohen’s Kappa coefficients metrics. Maximum accuracy and cohen’s Kappa coefficient of 0.7733 and 0.5488, respectively, were obtained using the Linear Discriminant Analysis (LDA) classifier. Finally, the proposed method was superior in performance and presents significant results in the alpha (α) frequency band and the combination of alpha (α) and beta (β). This leads to the conclusion that the proposed method is adequate for user intent identification in a motor imagery-based BCI system of users with no prior experience. |
|---|
