Limits...
A Generalizable Brain-Computer Interface (BCI) Using Machine Learning for Feature Discovery.

Nurse ES, Karoly PJ, Grayden DB, Freestone DR - PLoS ONE (2015)

Bottom Line: The classifier does not use extensive a-priori information, resulting in reduced reliance on highly specific domain knowledge.Instead of pre-defining features, the time-domain signal is input to a population of multi-layer perceptrons (MLPs) in order to perform a stochastic search for the best structure.Our new approach has been shown to give accurate results across different motor tasks and signal types as well as between subjects.

View Article: PubMed Central - PubMed

Affiliation: NeuroEngineering Laboratory, Department of Electrical & Electronic Engineering, The University of Melbourne, Melbourne, VIC, Australia, 3010; Centre for Neural Engineering, The University of Melbourne, Melbourne, VIC, Australia, 3010.

ABSTRACT
This work describes a generalized method for classifying motor-related neural signals for a brain-computer interface (BCI), based on a stochastic machine learning method. The method differs from the various feature extraction and selection techniques employed in many other BCI systems. The classifier does not use extensive a-priori information, resulting in reduced reliance on highly specific domain knowledge. Instead of pre-defining features, the time-domain signal is input to a population of multi-layer perceptrons (MLPs) in order to perform a stochastic search for the best structure. The results showed that the average performance of the new algorithm outperformed other published methods using the Berlin BCI IV (2008) competition dataset and was comparable to the best results in the Berlin BCI II (2002-3) competition dataset. The new method was also applied to electroencephalography (EEG) data recorded from five subjects undertaking a hand squeeze task and demonstrated high levels of accuracy with a mean classification accuracy of 78.9% after five-fold cross-validation. Our new approach has been shown to give accurate results across different motor tasks and signal types as well as between subjects.

No MeSH data available.


Related in: MedlinePlus

Performance of two-class hand squeeze classifier.The classification performance for the current method is shown as the average performance across the five-fold cross validation for each participant. A The average Cohen’s kappa score on unseen test sets. Error bars are standard deviation across 5 folds. The red dashed line shows chance performance (kappa = 0). White dashed lines indicate minor gridlines. The asterisks represent the confidence level to reject the  hypothesis that the results are not significantly different to chance performance (** for p < 0.01) using the kappa significance test (described in S1 Text). For the exact p-values see S2 Appendix. B The average percentage classification accuracy on the unseen test sets. The error bars are standard deviation after five-fold cross validation. The red dashed line represents chance outcome (50%). White dashed lines indicate minor gridlines.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4482677&req=5

pone.0131328.g005: Performance of two-class hand squeeze classifier.The classification performance for the current method is shown as the average performance across the five-fold cross validation for each participant. A The average Cohen’s kappa score on unseen test sets. Error bars are standard deviation across 5 folds. The red dashed line shows chance performance (kappa = 0). White dashed lines indicate minor gridlines. The asterisks represent the confidence level to reject the hypothesis that the results are not significantly different to chance performance (** for p < 0.01) using the kappa significance test (described in S1 Text). For the exact p-values see S2 Appendix. B The average percentage classification accuracy on the unseen test sets. The error bars are standard deviation after five-fold cross validation. The red dashed line represents chance outcome (50%). White dashed lines indicate minor gridlines.

Mentions: A plot of the cross-validated classification results for the two-class and three-class methods are shown in Figs 5 and 6, respectively. Across the five participants, the two-class classifier gave an average kappa of 0.58 and an average accuracy of 78.9%. The three-class classifier gave an average kappa of 0.37 and an accuracy of 60.7%. Both of these are well above the chance performance, which would give kappa values of zero for both methods. Figs 5 and 6 show that all the results were significantly above chance performance for all participants (p < 0.01) using the kappa significance test (described in S1 Text).


A Generalizable Brain-Computer Interface (BCI) Using Machine Learning for Feature Discovery.

Nurse ES, Karoly PJ, Grayden DB, Freestone DR - PLoS ONE (2015)

Performance of two-class hand squeeze classifier.The classification performance for the current method is shown as the average performance across the five-fold cross validation for each participant. A The average Cohen’s kappa score on unseen test sets. Error bars are standard deviation across 5 folds. The red dashed line shows chance performance (kappa = 0). White dashed lines indicate minor gridlines. The asterisks represent the confidence level to reject the  hypothesis that the results are not significantly different to chance performance (** for p < 0.01) using the kappa significance test (described in S1 Text). For the exact p-values see S2 Appendix. B The average percentage classification accuracy on the unseen test sets. The error bars are standard deviation after five-fold cross validation. The red dashed line represents chance outcome (50%). White dashed lines indicate minor gridlines.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4482677&req=5

pone.0131328.g005: Performance of two-class hand squeeze classifier.The classification performance for the current method is shown as the average performance across the five-fold cross validation for each participant. A The average Cohen’s kappa score on unseen test sets. Error bars are standard deviation across 5 folds. The red dashed line shows chance performance (kappa = 0). White dashed lines indicate minor gridlines. The asterisks represent the confidence level to reject the hypothesis that the results are not significantly different to chance performance (** for p < 0.01) using the kappa significance test (described in S1 Text). For the exact p-values see S2 Appendix. B The average percentage classification accuracy on the unseen test sets. The error bars are standard deviation after five-fold cross validation. The red dashed line represents chance outcome (50%). White dashed lines indicate minor gridlines.
Mentions: A plot of the cross-validated classification results for the two-class and three-class methods are shown in Figs 5 and 6, respectively. Across the five participants, the two-class classifier gave an average kappa of 0.58 and an average accuracy of 78.9%. The three-class classifier gave an average kappa of 0.37 and an accuracy of 60.7%. Both of these are well above the chance performance, which would give kappa values of zero for both methods. Figs 5 and 6 show that all the results were significantly above chance performance for all participants (p < 0.01) using the kappa significance test (described in S1 Text).

Bottom Line: The classifier does not use extensive a-priori information, resulting in reduced reliance on highly specific domain knowledge.Instead of pre-defining features, the time-domain signal is input to a population of multi-layer perceptrons (MLPs) in order to perform a stochastic search for the best structure.Our new approach has been shown to give accurate results across different motor tasks and signal types as well as between subjects.

View Article: PubMed Central - PubMed

Affiliation: NeuroEngineering Laboratory, Department of Electrical & Electronic Engineering, The University of Melbourne, Melbourne, VIC, Australia, 3010; Centre for Neural Engineering, The University of Melbourne, Melbourne, VIC, Australia, 3010.

ABSTRACT
This work describes a generalized method for classifying motor-related neural signals for a brain-computer interface (BCI), based on a stochastic machine learning method. The method differs from the various feature extraction and selection techniques employed in many other BCI systems. The classifier does not use extensive a-priori information, resulting in reduced reliance on highly specific domain knowledge. Instead of pre-defining features, the time-domain signal is input to a population of multi-layer perceptrons (MLPs) in order to perform a stochastic search for the best structure. The results showed that the average performance of the new algorithm outperformed other published methods using the Berlin BCI IV (2008) competition dataset and was comparable to the best results in the Berlin BCI II (2002-3) competition dataset. The new method was also applied to electroencephalography (EEG) data recorded from five subjects undertaking a hand squeeze task and demonstrated high levels of accuracy with a mean classification accuracy of 78.9% after five-fold cross-validation. Our new approach has been shown to give accurate results across different motor tasks and signal types as well as between subjects.

No MeSH data available.


Related in: MedlinePlus