.png){kind=icon}
.png){kind=icon}
Return
EEG based stereo vision assessment method using random dot stereograms with various motion patterns
.png){kind=icon}
PRE
AI
DOI：10.1016/j.bspc.2025.107531.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Stereoscopic vision is a complex visual function that allows humans to perceive depth, providing insights into three-dimensional perception. Studying brain activities during stereoscopic vision can enhance our understanding of its mechanisms and improve applications in medical diagnostics and entertainment. Traditional stereo vision tests, which primarily rely on static images, often fail to accurately reflect individual visual capabilities in real-world scenarios. To objectively and effectively assess individual stereoscopic vision capabilities under both dynamic and static conditions, we analyzed Electroencephalography (EEG) signals elicited under three distinct stereoscopic motion paradigms and developed a method named Channel-Frequency Feature Optimization Based on Random Forest (CFRF). Furthermore, we introduced the Multi-head Attention Long Short-Term Memory with Deep Residual Convolutional Network (MAL-DRCN). This network combines the Multi-head Attention Long Short-Term Memory (MAL) module for temporal feature extraction, which emphasizes the global significance of various time points, with the Long Short-Term Memory (LSTM) enhancing the temporal detail capture. The time-sensitive data is then processed by the Deep Residual Convolutional Network (DRCN) module, which uses a series of Residual Convolutional blocks (RCN) to extract complex spatiotemporal features and address the gradient vanishing issue effectively. In trichotomous classification tasks that involved crossed and uncrossed disparities, our method achieved impressive accuracies of 95.27% and 94.44%, respectively.
Keywords:
EEG
Stereo vision
Random dot stereogram
Random forest
Channel-frequency feature optimization
Journal
IF:
4.9
Papers:
9.8K
Citations:
2.4W