知识驱动的SSVEP刺激界面序贯式实验方案优化

To solve the problems of large number of experiments, high cost and long cycle in the traditional experimental design method when designing the stimulus interface of steady state visual evoked potential (SSVEP), a knowledge-driven method for optimizing the sequential experimental scheme of SSVEP stimulus interface was proposed. Taking SSVEP stimulus interface parameters as design variables and response performance as optimization objective, the initial sample space was built. The prior knowledge of the stimulus interface parameters was characterized in terms of probability models, and the warped sample space was reconstructed with probability integral transformation, which narrowed the region with low probability of optimum value and expanded the region with a high probability of optimum value. The expected improved acquisition function was used for iterative optimization to obtain the optimal stimulus interface with less experiment times. The experimental results indicated that the proposed optimization method could reduce the number of experiments by about 53% and 44%, respectively, compared with the Latin hypercubic and orthogonal design methods under the premise of guaranteeing the best optimal stimulus parameters.