Biomedical Engineering Reference
In-Depth Information
color/gray gratings were used as stimuli, each having one of four orientations (0,
45, 90, 135 degrees) and one of four colors (red, blue, green, yellow). To ensure
that response differences were due to color and orientation only, all gratings were
isoluminant at 10 cd/m 2 . Each grating was presented 32 times, and was thus associ-
ated with 32 single-trial response images. Figure 10.2 shows the averaged response
images that were associated with four different stimuli. This figure demonstrates
that each stimulus gave rise to a unique pattern of responses across the imaged area.
Figure 10.3 Decoding the stimulus color from a single-trial response. All single-trial responses
except those from the last imaging block were grouped according to stimulus color, regardless
of their orientations. These four groups of responses were used to train a set of SVM classifiers,
which were then used to decode the stimulus color of each single-trial image from the last block.
This procedure was repeated 32 times, each time using a different block of data, to test the SVM
classifiers after they have been trained by the remaining data. The decoded stimulus color was
compared with the real one for each testing response, and the percentage of correct decoding
across the entire dataset represents a quantitative measure of the average amount of information
about color that was encoded in a single-trial response.
Search WWH ::




Custom Search