Decoding Phase-Based Information from Steady-State Visual Evoked Potentials with Use of Complex-Valued Neural Network

摘要

In this paper, we report on the decoding of phase-based information from steady-state visual evoked potential (SSVEP) recordings by means of a multilayer feedforward neural network based on multivalued neurons. Networks of this kind have inputs and outputs which are well fitted for the considered task. The dependency of the decoding accuracy w.r.t. the number of targets and the decoding window size is discussed. Comparing existing phase-based SSVEP decoding methods with the proposed approach, we show that the latter performs better for the larger amount of target classes and the sufficient size of decoding window. The necessity of the proper frequency selection for each subject is discussed.