Rehabilitation Engineering
Rahele Mohammadi; Ali Mahloojifar
Volume 7, Issue 1 , June 2013, , Pages 39-55
Abstract
A critical issue in designing a self-paced brain computer interface (BCI) system is onset detection of the mental task from the continuous electroencephalogram (EEG) signal to produce a brain switch. This work shows significant improvement in a movement based self-paced BCI by applying a new sparse learning ...
Read More
A critical issue in designing a self-paced brain computer interface (BCI) system is onset detection of the mental task from the continuous electroencephalogram (EEG) signal to produce a brain switch. This work shows significant improvement in a movement based self-paced BCI by applying a new sparse learning classification algorithm, probabilistic classification vector machines (PCVMs) to classify EEG signal. Constant-Q filters instead of constant bandwidth filters for frequency decomposition are also shown to enhance the discrimination of movement related patterns from EEG patterns associated with idle state. Analysis of the data recorded from seven subjects executing foot movement using the constant-Q filters and PCVMs shows a statistically significant 16% (p<0.03) average improvement in true positive rate (TPR) and a 2% (p<0.03) reduction in false positive rate (FPR) compared with applying constant bandwidth filters and SVM classifier.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Rahele Mohammadi; Ali Mahloojifar
Volume 6, Issue 2 , June 2012, , Pages 141-152
Abstract
Self-paced BCI systems are more natural for real-life applications since these systems allow the user to control the system when desired. Detection of event periods in continuous EEG signal is one of the most important challenges in designing self-paced BCIs. In this paper, the Event related synchronization ...
Read More
Self-paced BCI systems are more natural for real-life applications since these systems allow the user to control the system when desired. Detection of event periods in continuous EEG signal is one of the most important challenges in designing self-paced BCIs. In this paper, the Event related synchronization (ERS) is extracted from idle EEG signal using fractal dimensions in frequency range from 6 to 36 Hz and sparse representation based classifier. Our proposed method applied on EEG signal recorded during executing foot movement in 7 subjects. The average true positive rate and false positive rate equal to 90% and 5% were achieved.
