Vision mechanism model using brain–computer interface for light sensing

The electroencephalograph (EEG) learning network model (EEGNet) is developed according to the convolutional neural network model architecture. It can be applied in the area of the EEG recognition because the EEGNet has the advantage of adapting to the EEG processing. However, the application has a bottleneck problem that the EEG selection of the specific brain–computer interface (BCI) affects the EEGNet recognition accuracy. In this paper, we developed an integrated EEGNet model of the human vision mechanism for light intensity perception. First, the special BCI is constructed by using a designed multiplexer, the EEG acquisition circuit, the magnifier and the filter. Second, the effect of the underground environment illumination on EEG is explored by using the constructed BCI. Third, the model of the vision mechanism is established by using the integrated EEGNet. Finally, experiments show that the integrated EEGNet increases the light intensity recognition accuracy respectively by 8.4% and 3.9%, compared with the multi-channel EEGNet and the single channel EEGNet. The integrated EEGNet effectively perceives and recognizes the underground illumination intensities, dim intensity of 0–60 Lx, mild intensity of 61–120 Lx, and bright intensity of 121–350 Lx. The proposed model can provide useful references for miner helmet or other special environment light-related devices.