We have created a BCI inspired by the game of Clue. Rather than asking what the opponent has in their envelope, the BCI reads what the subject has in their mind. In our role-play scenario, the subject has committed a crime and their accomplice is on the loose. This accomplice is known as a "Probe"--hidden information that only the subject knows, and the computer is trying to learn.
During each trial, python scripts gather data from the OpenBCI cyton board. Our player is shown a series of faces via a Pygame GUI and another python script synchronizes presented stimulus with recorded brain activity. All displayed faces are unfamiliar except for the Probe (unknown to experimenters) and Target image. The player is asked to press a button when the Target appears to ensure that they are paying attention.
The data is filtered and detrended, then the P300 elicited by each image flash is identified and labeled as either Target or Unknown. The inherent salience of the Target and Probe stimuli should result in similarities between their P300s. At this point, the Probe is hidden amongst the unknowns. We use a bootstrap correlation analysis to determine which of the Unknown faces is most similar to the Target – this face is the Probe.
Log in or sign up for Devpost to join the conversation.