<p>Virtual Reality Motion Sickness (VRMS) seriously affects the user experience and health in virtual reality, and is one of the urgent issues to be solved in the virtual reality industry. Accurate detection is the prerequisite for studying the causes and mitigation methods of VRMS. Therefore, this paper proposes a new EEG-based model for VRMS detection. The current model tends to ignore the differences between leads in different brain regions when processing multi-lead EEG, and fuses multi-lead signals prematurely in the feature extraction stage. The model proposed in this paper combines Lead Embedding (LE) and Siamese Network (SN). The LE in the model first encodes the lead of EEG into the codes and adds it to the EEG of each lead. The multi-lead EEG is then split into individual lead signals and sent to the SN respectively. The SN uses LSTM to extract the time features of each lead. Finally, the extracted features of all leads are stacked together and classified. A simulated flight scene is used to induce VRMS, and the EEG of the subjects in the resting state before and after the induced task is collected to verify the proposed model. The accuracy of the proposed model for VRMS recognition reaches 98.8%. The results show that the proposed model can be used to detect VRMS and provide a new idea for processing multi-lead EEG and extracting its spatial and temporal information. This method is expected to aid in eliminating user motion sickness and optimizing VR technology.</p>