A Study On P300-Based Brain-Computer Interface Speller

Brain-computer interface(BCI) does not depend on the organization of the muscles by directly translating recorded electrical brain signals into corresponding machine commands to control external devices, so as to complete the process of human-computer interaction. Along with the development of the information technology, BCI is also becoming a hot topic. It has set up an independent communication channel between the human brain and external environment, and provides a way to communicate with the outside world for people with serve motor disabilities. In addition, the research of BCI is applied in many fields, such as medicine, controlling, artificial intelligence, brain science and so on. In this thesis, we study the BCI system based on P300, which is one of the most common BCI applications. Although there are many research achievements regarding P300-based BCI, in practical applications the system still has low transmission rate and needs idle state detection in asynchronous mode. Facing these problems, this thesis made some progress in the following two aspects:First, a method of P300 detection is designed,which is named stimulation sequence reduction based on dynamic stopping criterion(DSC).The algorithm based on Bayesian framework is used to calculate the posterior probability of each character becoming target character, then the probability value is compared with the threshold.If it is greater, the flashing sequence stops and the target character is directly output;otherwise it will be sorted according to probability value and deletes some characters of smaller probability values, hence reducing the interference of the target character. The experimental results show that the speed of the input characters is significantly improved while the accuracy is basically not declined.Secondly, this thesis discusses the character input system of asynchronous P300 based on BCI. The system can detect the P300 signal whether the user is typing. Hence, there exists an idle state compared to synchronous P300 system. For the detection of control state/idle state, this thesis introduces two kinds of methods: one is a computational approach based on statistical analysis. The algorithm based on Bayesian framework is used to calculate posterior probability of each flashing sequence, then the probability value is compared with the threshold. If it exceeds the threshold, the control state is judged and the target character based on DSC is output. The other one is based on the combined classifier, which establishes a classifier model of two levels and judges the state according to the label. If the label is one, the control state will be judged and the target character based on DSC is output. By analyzing the experimental data, the results show that two states of the asynchronous mode can be accurately detected by 90% in the system.