Abstract
Recent evidence suggests that visual perception is a rhythmic process: target detection undergoes periodic fluctuations at frequencies from 7 to 15Hz, and these rapid changes can be linked to the phase of ongoing brain oscillations. In turn, these oscillations are modulated by visual stimulation: for example, the impulse response function during a stream of random (white-noise) luminance changes includes a strong oscillatory component at ~10Hz akin to a perceptual echo. The phase, amplitude and duration of this ‘echo function’ are observer-specific. Thus, the same white-noise sequence produces markedly different but predictable patterns of oscillatory responses in different observers. We combined these observations to establish the feasibility of a ‘neuro-encryption’ system. We first recorded the EEG impulse response (‘echo’) function of each individual observer (N=9). In a separate session without EEG, we embedded threshold-intensity targets in streams of white-noise luminance fluctuations; all subjects viewed the same sequences with the same targets. For each observer, we used the previously-recorded echo function to estimate (by convolution with the white noise sequence) the oscillatory brain response around each target. We then compared oscillatory phase for detected and undetected targets, and found a significant opposition between them: around a certain phase the targets were more likely (by ~10%) to be detected than at the opposite phase. Importantly, this performance difference persisted when considering only trials for which two subjects had opposite phase estimates: the observer around the ‘good’ phase performed better (paired t-test, p< 0.001) than the one at the ‘bad’ phase, for the same target embedded in the same random sequence. In sum, for a given subject we can predict target detection in any random sequence, without concurrently recording brain signals. This augurs the possibility of creating stimulus sequences in which only one subject can perceive a target message, a form of ‘neuro-encryption’.
Meeting abstract presented at VSS 2015