Research On Multi-target Coding Technology For SSVEP-based Brain-computer Interface

Brain-computer interface system(BCIs)based on steady-state visual evoked potential(SSVEP)is a mainstream brain-computer interface system.It can achieve non-invasive and high information transfer rate(ITR)brain-computer direct communication with low software/hardware cost,so it has become a research hotspot in BCI field.Due to human physiological mechanism,SSVEP BCI,especially those with medium or high frequency stimulus,usually cannot provide many target instructions.To solve this problem,this paper studies from two aspects of paradigm coding optimization and training cost optimization,and provides a feasible technical method and implementation scheme for multi-target high-frequency SSVEP-BCI application.In terms of target coding,based on the idea of time-division coding,this paper proposes a multi symbol time division coding(MSTDC),as well as response difference maximization coding criterion and optimization algorithm,which can achieve any number of codeword target coding with finite symbol elements.By implementing this method,a high-frequency SSVEP-BCI system with multiple targets is designed and implemented.The system adopts time-division phase coding mode,and uses 0,?/2,?,3?/2 phase-coded stimuli as symbols at a single frequency of 30 Hz to achieve 40 candidate codewords.In the online experiment with 12 subjects,the system achieved an average recognition accuracy of 96.77%±2.47%and an average ITR of 119.05±6.11 bits/min.To improve the training efficiency,this paper further proposes a Template-splicing TRCA(TS-TRCA)algorithm.The algorithm can achieve the dynamic expansion of the number of candidate targets without repeated training.Through this method,72 targets are detected and recognized with 40-target training dataset,and average recognition accuracy of 86.23%±7.75%were achieved,resulting in average ITR of 95.68±14.19 bits/min.To further reduce the training cost,this paper applies MSTDC to design a less training SSVEP-BCI system.The SSVEP-BCI system adopts the training mode of single target stimulation to generate training templates of all symbols,which greatly simplifies the SSVEP-BCI training process.Based on TS-TRCA,this paper further presents a symbol-TRCA algorithm for the less training system.In the online experiment with 40 high frequency targets,each subject only spent 36s on training,achieved an average ITR of 97.41±19.12 bits/min.The results of the experiment show that the system is a SSVEP-BCI system with the advantages of multi-target,easy training and high-comfortability.