Enhancing Performance Of P300-Speller Under Effect Of Mental Workload

As one of the major BCI paradigms,P300-Speller has been extensively studied in recent years attributed to its advantages such as no need for training and good stability.However,in practical application of P300-Speller,users may experience various mental states,such as changes of mental workload,and this may lead to impairement of performance.Therefore,developimg effective methods of mitigating impact of mental workload and improving performance in practical uses are of great significance.In this study,with dual-task paradigm stimulating practical situation,methods from two aspects,offline training model and online stopping criteria,were developed to improve performance of P300-Speller under dual-task paradigm.For the method of building new training model,the spelling task of P300-Speller was carried out concurrent with n-back or mental arithmetic task,and speller-only task was performed as the control condition.Data under dual-task conditions were introduced to speller-only data respectively to build new training models of P300-Speller classifier,and tested under the two dual-task conditions to investigate the effectiveness of the new models.This study was designed to propose a new method to mitigate the negative effects of mental workload on P300-Speller and verify its effectiveness.The results showed that the mixed-data training models were effective in improving the performance of P300-Speller under dual-task conditions,even if the task of imported training data is different from the task of testing data.The research findings in this study confirmed the feasibility of building a universal training model applied in real-time operation of P300-Speller to enhance its performance under mental workload in practical situations.For the method of importing dynamic stopping criteria,the spelling task of online P300-Speller was carried out concurrent with 2-back and 3-back task,with speller-only task serving as the control condition.In real-time signal processing,dynamic stopping criteria was adopted to control the number of flash repetitions automatically according to recognition results,and the condition with static stopping criteria was performed as the control condition.Then performance under different task conditions and stopping stratigies were evaluated and compared by indicators such as information transfer rate,character recognition accuracy and spelling time.The results showed that information transfer rate under varied mental workload condition has been significantly improved with dynamic stopping criteria.Besides,by this strategy flash repetition number increased with rising mental workload level to ensure high recognition accuracy and stable performance of P300-Speller.These results confirmed the effectiveness of dynamic stopping criteria in improving P300-Speller performance under various mental workload conditions.Aiming at the problem of sensitivity of P300-Speller performance to mental workload,methods of mixed-data training model and dynamic stopping criteria were proposed in this study.The results under dual-task paradigm confirmed the effectiveness of these methods in mitigating influence of mental workload and improve performance of P300-Speller under dual-task condition,which are of great significance to improving robustness of P300-Speller.