Effective classification of motor imagery electroencephalograph (EEG) data is an important challenge. Spatial filtering such as Common Spatial Pattern (CSP) and its variants are commonly used for this task. However, CSP effectiveness depends on the subject-specific frequency band. Even by optimally selecting a subject-specific frequency band, this method still fails for some subjects. On the other hand, some studies suggest that temporal features may discriminate classes more efficiently. This work proposes a hybrid method based on elastic net and Least Absolute Shrinkage and Selector Operator (LASSO) to optimally select between spatial and temporal features. This algorithm uses joint spatial and temporal features followed by an optimal combined feature selection scheme for each subject. Results show significant improvement for subjects whose spatial features failed to produce acceptable results and overall improvement over the combined data.
Classification of EEG signals using the Spatio-Temporal feature selection via the elastic net / Noei, S.; Ashtari, P.; Jahed, M.; Vahdat, B. V.. - (2017), pp. 232-236. (Intervento presentato al convegno 23rd Iranian Conference on Biomedical Engineering and 2016 1st International Iranian Conference on Biomedical Engineering, ICBME 2016 tenutosi a irn nel 2016) [10.1109/ICBME.2016.7890962].
Classification of EEG signals using the Spatio-Temporal feature selection via the elastic net
Noei S.;
2017-01-01
Abstract
Effective classification of motor imagery electroencephalograph (EEG) data is an important challenge. Spatial filtering such as Common Spatial Pattern (CSP) and its variants are commonly used for this task. However, CSP effectiveness depends on the subject-specific frequency band. Even by optimally selecting a subject-specific frequency band, this method still fails for some subjects. On the other hand, some studies suggest that temporal features may discriminate classes more efficiently. This work proposes a hybrid method based on elastic net and Least Absolute Shrinkage and Selector Operator (LASSO) to optimally select between spatial and temporal features. This algorithm uses joint spatial and temporal features followed by an optimal combined feature selection scheme for each subject. Results show significant improvement for subjects whose spatial features failed to produce acceptable results and overall improvement over the combined data.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione
