THE GOVERNING COUNCIL OF THE UNIVERSITY OF TORONTO
发明人:
Adrian Razvan NESTOR,Dan NEMRODOV
申请号：
US16423064
公开号：
US20190357797A1
申请日:
2019.05.27
申请国别(地区):
US
年份:
2019
代理人:
摘要：
There is provided a system and method for generating visual category reconstruction from electroencephalography (EEG) signals. The method includes: receiving scalp EEG signals； using a trained pattern classifier, determining pairwise discrimination of the EEG signals, the pattern classifier trained using a training set comprising EEG signals associated with a subject experiencing different known visual identities； determining discriminability estimates of the pairwise discrimination of the EEG signals by constructing a confusability matrix； generating a multidimensional visual representational space； determining visual features for each dimension of the visual representational space by determining weighted sums of image stimulus properties； identifying subspaces determined to be relevant for reconstruction； and reconstructing the visual appearance of a reconstruction target using estimated coordinates of the target and a summed linear combination of the visual features proportional with the coordinates of the target in the visual representational space.
