| Brain-computer interfaces(BCIs) can utilize the electrical signals generated by the central nervous system(CNS), and then translate these signals into users’ commands, which can control some external systems, such as personal computer, electrically propelled wheelchairs and prosthesis. BCI system can provide a new bridge between the brain and the world that bypass the human body. Therefore, this device can be used as an alternative communication method to improve the life-quality of people with amyotrophic lateral sclerosis(ALS), stroke and some other neuromuscular diseases and may change the way we are communicating, entertaining and living remarkably in the future.The feature extraction of event-related potentials(ERPs) is a significant prerequisite for Imitating-Reading BCI system. In this study, we proposed an adaptive multi-ganglion artificial neural network based feature learning(ANNFL) method to extract feature structures of single-trial multi-channel ERP signals and improve classification accuracies. Seven subjects took part in the Imitating-Reading ERP experiments. We recorded the target electroencephalography(EEG) samples(elicited by target stimuli), non-target samples(elicited by non-target stimuli) and the EEG samples in reading, relaxation and closing-eye statuses for each subjects. Then we applied ANNFL method to extract the feature vectors and classified them by using support vector machine(SVM). The ANNFL method outperforms the common spatial pattern(CSP) method, then leads to higher classification accuracies of seven subjects’ BCI signals than using the single-channel temporal features. ANNFL is an unsupervised feature learning method, which can automatically learn feature vector from EEG data and provide more effective feature representation than CSP method and single-channel temporal feature extraction method. |
