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

Temporal filtering of artificial neural networks (ANNs).Subplots are the estimated power spectral densities (PSDs) from the periodogram of the temporal weights. Temporal weights are taken between the input layer and the first hidden layer averaged over space (channels) and across the neurons in the first hidden layer to obtain a single value for each point in time. The initial weight values were obtained from the EEG epoch corresponding to each channel, with the DC component removed. The solid line represents the mean PSD over five folds of cross validation. The shaded region represents the distance between the minimum and maximum obtained values. A Subject A. B Subject B. C Subject C. D Subject D. E Subject E.
© Copyright Policy
Related In: Results  -  Collection

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

pone.0131328.g010: Temporal filtering of artificial neural networks (ANNs).Subplots are the estimated power spectral densities (PSDs) from the periodogram of the temporal weights. Temporal weights are taken between the input layer and the first hidden layer averaged over space (channels) and across the neurons in the first hidden layer to obtain a single value for each point in time. The initial weight values were obtained from the EEG epoch corresponding to each channel, with the DC component removed. The solid line represents the mean PSD over five folds of cross validation. The shaded region represents the distance between the minimum and maximum obtained values. A Subject A. B Subject B. C Subject C. D Subject D. E Subject E.

Mentions: Fig 10 is an investigation of which frequencies in the EEG spectrum the ANNs emphasize. Emphasis is inferred using the average weight values between the input signal and each neuron in the first hidden layer, which are averaged over space and across the neurons in the first hidden layer. An interpretation of these weights is as finite impulse response filter coefficients acting on the input signal. Individual neurons may emphasize different spectral information and it is by using many of these individual filters that the ANN is able to construct complex features from high-dimensional data. It can be seen from Fig 10 that on average the lower frequencies appear to be accentuated by all the subjects ANNs. Although it is worthwhile to note that the analyses were averaged over many individual filters and do not represent temporal or spectral features created by the ANNs in and of themselves.


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

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

Temporal filtering of artificial neural networks (ANNs).Subplots are the estimated power spectral densities (PSDs) from the periodogram of the temporal weights. Temporal weights are taken between the input layer and the first hidden layer averaged over space (channels) and across the neurons in the first hidden layer to obtain a single value for each point in time. The initial weight values were obtained from the EEG epoch corresponding to each channel, with the DC component removed. The solid line represents the mean PSD over five folds of cross validation. The shaded region represents the distance between the minimum and maximum obtained values. A Subject A. B Subject B. C Subject C. D Subject D. E Subject E.
© Copyright Policy
Related In: Results  -  Collection

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

pone.0131328.g010: Temporal filtering of artificial neural networks (ANNs).Subplots are the estimated power spectral densities (PSDs) from the periodogram of the temporal weights. Temporal weights are taken between the input layer and the first hidden layer averaged over space (channels) and across the neurons in the first hidden layer to obtain a single value for each point in time. The initial weight values were obtained from the EEG epoch corresponding to each channel, with the DC component removed. The solid line represents the mean PSD over five folds of cross validation. The shaded region represents the distance between the minimum and maximum obtained values. A Subject A. B Subject B. C Subject C. D Subject D. E Subject E.
Mentions: Fig 10 is an investigation of which frequencies in the EEG spectrum the ANNs emphasize. Emphasis is inferred using the average weight values between the input signal and each neuron in the first hidden layer, which are averaged over space and across the neurons in the first hidden layer. An interpretation of these weights is as finite impulse response filter coefficients acting on the input signal. Individual neurons may emphasize different spectral information and it is by using many of these individual filters that the ANN is able to construct complex features from high-dimensional data. It can be seen from Fig 10 that on average the lower frequencies appear to be accentuated by all the subjects ANNs. Although it is worthwhile to note that the analyses were averaged over many individual filters and do not represent temporal or spectral features created by the ANNs in and of themselves.

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