MEGA: A Multimodal EEG-Based Visual Fatigue Assessment System

SSVEP-BCI are renowned for their high precision, rapid data transmission, and the advantage of not requiring user training. However, the intense brightness, frequent lowfrequency stimuli, and monotony of tasks required for SSVEP signal generation can lead to visual fatigue. This paper addresses the issue of visual fatigue within SSVEP-based BCI systems. We introduce a real-time fatigue assessment system named MEGA, which features a high-precision, wearable EEG device engineered with ARM architecture. The system employs a multimodal fusion classification algorithm, enhancing the comfort and performance of the SSVEP-BCI system during operation. By analyzing characteristics from EEG channels O1, O2, Fp1, and Fp2, extracting blink features using STFT, and integrating EEG entropy, we have explored a robust set of fatigue classification features. This streamlined and sensitive feature combination reduces system complexity while improving classifier sensitivity. The proposed multimodal transformer architecture for fatigue classification aims to enhance both the comfort and performance during SSVEP-BCI system operation. Ultimately, the proposed system achieved a classification accuracy of 98.89%, which is 4.89% higher than the latest visual fatigue classification systems.