Detecting EEG evoked responses for target image search with mixed effect models

There is evidence that brain signals associated with perceptual processes can be used for target image search. We describe the application of mixed effect models (MEMs) to brain signature detection. We develop an MEM detector for detecting brain evoked responses generated by perceptual processes in the human brain associated with detecting novel target stimuli. We construct the model using principal component analysis and linear discriminant analysis (LDA) bases. We adopt the LDA for dimension reduction. For parameter regularization we use 10-fold cross validation and report experimental results from six subjects. Four out of six subjects achieve very good detection performance with more than 0.9 areas under receiver operating characteristic curves. The results demonstrate that the MEM can provide reliable inference on single-trial ERP detection on the task of target image search.