关键词: 稳态视觉诱发电位, 脑-机接口, 多码元, 时分编码, 少训练

Abstract: Steady-State Visual Evoked Potentials (SSVEP) Brain Computer Interface (BCI) is one of the three mainstream paradigms in the field of noninvasive brain-computer interface research. Existing studies have shown that the recognition efficiency of SSVEP-BCI can be effectively improved by introducing the calibrated data of the subjects. However, the traditional SSVEP-BCI training algorithm have to collect a large number of subjects' EEG data before training, which greatly increases the cost of SSVEP-BCI, thus disadvantage to the popularization of SSVEP-BCI. Therefore, how to reduce the training cost of the system meanwhile ensuring the efficiency of BCI recognition has become one of the research hotspots of SSVEP-BCI. In order to reduce the defect of long training on SSVEP-BCI., this study proposes a less-training detection algorithm based on multi-symbol time-division coding for SSVEP-BCI. The algorithm makes full use of the advantages of the multi-symbol time-division coding scheme, which uses the symbol response multiplexing method to identify a large number of candidate targets with a small amount of EEG training data. In the online experiment of 40 target 30Hz high frequency encoded brain-computer interface system with 12 subjects, the algorithm can achieve an average recognition accuracy of 86.04%±10.27% and information transfer rate of 97.41±19.12bits/min with only 36s training time. The experimental results show that the algorithm can achieve high recognition accuracy and information transmission rate with a small amount of training data, thus it is expected to improve the practical application value of SSVEP-BCI.

Key words: steady-state visual evoked potentials, brain-computer interface, multi-symbol, time division coding, less training