2015
DOI: 10.1016/j.neucom.2015.02.005
|View full text |Cite
|
Sign up to set email alerts
|

Adaptive semi-supervised classification to reduce intersession non-stationarity in multiclass motor imagery-based brain–computer interfaces

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

1
44
0
1

Year Published

2016
2016
2024
2024

Publication Types

Select...
5
3
1

Relationship

2
7

Authors

Journals

citations
Cited by 77 publications
(46 citation statements)
references
References 50 publications
1
44
0
1
Order By: Relevance
“…The CSP algorithm is an efficient feature extraction algorithm that has been extensively used in MI-based BCI systems [45][46][47][48][49][50]. The CSP is based on the simultaneous diagonalization of two covariance matrices.…”
Section: Common Spatial Patternmentioning
confidence: 99%
“…The CSP algorithm is an efficient feature extraction algorithm that has been extensively used in MI-based BCI systems [45][46][47][48][49][50]. The CSP is based on the simultaneous diagonalization of two covariance matrices.…”
Section: Common Spatial Patternmentioning
confidence: 99%
“…This behavior is caused by an increase of the number of users that could not finish the task, likely due to the intersession variability of the EEG, which cannot be followed by the constant threshold. Those nonstationary changes of the EEG are emphasized as more sessions are carried out without updating the custom classifier of each user and actually constitute one of the main limitations of the current BCI systems [29], [30].…”
Section: Discussionmentioning
confidence: 99%
“…As previously indicated, the major drawback of this kind of applications is the classifier performance variability between sessions and users. Reducing this variability and increasing the classification accuracy by using more suitable processing techniques in both feature extraction and selection could improve the robustness of the system [29], [30]. In addition, control state threshold is calculated directly over the SWLDA scores and thus, it depends on the classifier performance of each user.…”
Section: Discussionmentioning
confidence: 99%
“…Este grupo ha desarrollado estudios orientados a caracterizar la actividad EEG basal mediante métricas derivadas, por ejemplo, de la teoría de redes complejas (Lubeiro et al, 2015); así como orientadas a mejorar el procesado de las interfaces cerebro-computadora (Corralejo et al, 2011;Nicolas-Alonso, 2015;Martínez-Cagigal et al, 2017). Con respecto al procesado de la señal MEG, el GIBUVa ha tratado de caracterizar los cambios que se producen a causa del envejecimiento (Fernández et al, 2012;Gómez et al, 2013) y ayudar al diagnóstico de diferentes patologías, como la enfermedad de Parkinson (Gómez et al, 2011); así como mejorar el procesado de la señal relativo a la cancelación de artefactos (Escudero et al, 2011) y reducir el coste computacional en el análisis MEG (Martínez-Zarzuela et al, 2013).…”
Section: Procesado De Eeg Y Megunclassified