Introduction:The non-stationarity of electroencephalograms(EEG)has a substantial effect on the performance of classifiers in brain-computer interface(BCI)systems.To tackle this challenge,an adaptable version of the li...Introduction:The non-stationarity of electroencephalograms(EEG)has a substantial effect on the performance of classifiers in brain-computer interface(BCI)systems.To tackle this challenge,an adaptable version of the linear discriminant analysis(LDA)classifier was proposed.Accuracy is crucial when using BCIs for neurorestorative tasks or memory improvement.The accurate comprehension of brain responses facilitates more focused interventions,which may improve neurorestorative outcomes.BCIs equipped with adaptive classifers are useful for personalizing therapies to individual requirements and for improving neurorestorative processes.Notably,neurorestorative interventions that yield consistent,accurate,and reliable outcomes are more likely to inspire trust and elicit satisfaction in users.Methods:The proposed classifier continuously modified its parameters in accordance with EEG signals.The covariance matrix and mean values for each pair of classes were the updating parameters.The proposed classifier modifed the model parameters by prioritizing current signal data over older signal history.The proposed classifier was tested using a hybrid SSVEP+P300 BCI system.Results and conclusions:The proposed classifier had an estimated classification accuracy of 97.4%,and was more effective than pooled mean LDA and conventional multiclass LDA classifiers.Increased classification accuracy may increase the responsiveness of neurorestorative interventions and increase the usefulness of BCIs in neurorestoration.展开更多
文摘Introduction:The non-stationarity of electroencephalograms(EEG)has a substantial effect on the performance of classifiers in brain-computer interface(BCI)systems.To tackle this challenge,an adaptable version of the linear discriminant analysis(LDA)classifier was proposed.Accuracy is crucial when using BCIs for neurorestorative tasks or memory improvement.The accurate comprehension of brain responses facilitates more focused interventions,which may improve neurorestorative outcomes.BCIs equipped with adaptive classifers are useful for personalizing therapies to individual requirements and for improving neurorestorative processes.Notably,neurorestorative interventions that yield consistent,accurate,and reliable outcomes are more likely to inspire trust and elicit satisfaction in users.Methods:The proposed classifier continuously modified its parameters in accordance with EEG signals.The covariance matrix and mean values for each pair of classes were the updating parameters.The proposed classifier modifed the model parameters by prioritizing current signal data over older signal history.The proposed classifier was tested using a hybrid SSVEP+P300 BCI system.Results and conclusions:The proposed classifier had an estimated classification accuracy of 97.4%,and was more effective than pooled mean LDA and conventional multiclass LDA classifiers.Increased classification accuracy may increase the responsiveness of neurorestorative interventions and increase the usefulness of BCIs in neurorestoration.
