Fig 5. Overall process of the proposed model framework.

(a) Data Collection and Processing: Primarily, meticulous acquisition and processing of EEG signals and questionnaire data lay the foundation for this method. Raw data undergoes a transformational journey to become structured feature vectors conducive to computational analysis. Preprocessing involves the application of the wavelet packet transform [34] to EEG data, stratifying it into δ, θ, α, and β frequency components. This nuanced stratification aids in comprehending evolving EEG features over time, enhancing the model’s predictive accuracy. (b) Model Training: The training phase involves feeding preprocessed EEG signals and questionnaire data into the model’s architecture. A Recurrent Neural Network (RNN) takes charge, extracting distinct features characterizing different states. These features, essential for subsequent modeling, are then mapped to questionnaire responses, enabling the model to discern motion sickness states. (c) Test Data Evaluation: The final step involves rigorous evaluation using a classifier and predictor to assess the model’s performance on test data. This phase offers insights into the model’s generalization ability and predictive accuracy.