EEG Signal Recognition of VR Education Game Players Based on Hybrid Improved Wavelet Threshold and LSTM

According to Electroenc-Encephalo-Graphy (EEG) signal correlation analysis, attention levels of game players can objectively reflect changes in attention during Virtual Reality (VR) education games. To avoid noise interference, denoising processing is necessary. This study improves the traditional wavelet thresholding method by combining it with Ensemble Empirical Mode Decomposition (EEMD). Then, feature extraction is performed to remove noise and identify EEG features related to attention, which are classified using long short-term memory techniques. The proposed method is validated through experimental design, showing minimal root mean square error of 12.0231 and maximum signal-to-noise ratio of 11.3272, indicating effective denoising. In VR, attention is more focused and stable compared to the 3D environment. Time required to achieve challenge goals is 53.65 seconds in VR and 65.7 seconds in 3D environments, suggesting participants achieve goals earlier in VR. The correlation coefficient between VR and 3D environment is 0.784, indicating a strong correlation, with a Significant Difference (SD) in time required to achieve initial goals. The proposed method demonstrates effectiveness in EEG signal recognition