The Design And Implementation Of Asynchronous BCI Systems Based On Readiness Potentials

Asynchronous brain computer interface (BCI) provides a substitutive motor output pathways without peripheral neural system, and it has great value in motor assisting and neurological rehabilitation. Many laboratories have developed asynchronous BCI system, but due to the low real-time, recognition t small task and etc, most of the asynchronous BCI are used in laboratory envionments. We extract the characteristics of the motor preparation potential and predict the motion intention, then send out the control instructions in advance and improve the real-time performance of the system.We design an asynchronous BCI system based on motor preparation, which provides a new idea for asynchronous BCI system implementation.We design the asynchronous BCI system, in the complex environment in order to BCI system control and the way to complete the control task. The main task is as follows:(1) We designed an experimental paradigm for asynchronous BCI systems based on motor preparation potentials.We designed the experimental paradigm of complete autonomic movement in order to avoid any stimulation of the brain electrical signal.We set the fit of the single experimental time length, thus avoiding long time experiment of subjects caused by wmotional problems and cause the signal to drop, but not because of the large amount of data and increase the difficulty of data processing.We present our approach to the real time problem of asynchronous BCI systems.We through off-line motion analysis to EEG signal process characteristic by drawing a brain topographic map method and localization of activated brain regions and select the key channel, signal analysis frequency domain characterstics, determine the sensitive frequency and frequency segments.(2) Design control online experiments, Lu Wei robot, brain electrical signal is decoded as the control command, through the local area network control Lu Wei robot movement to the left or right direction, and ultimately through the EEG signals control Lu Wei robot avoiding obstacles on the path, realize the real-time control of parents machine. According to the feature of wavelet packet decomposition, support vector machine is used to classify and identify, and the average correct rate of prediction is 77.22%.The starting time of the motion prediction in motion before the start of the 750+30ms, the CSP algorithm extracts the relevant features, classification by support vector machine, seek to predict the optimal time window length predicted movement with the intention of the average correct rate 83.15%～92.71%.(3) We designed the control system to control the robot’s on-line test, through the EEG signal decoding to control commands, and then through the local area network to control the robot’s left and right motion direction, through the EEG signals control robot avoiding obstacles on the path, realize the real-time control of robot parents. Online test to prediction average of 874ms and prediction average correct rare for 79.18%.