The Importance of the Signal/Noise Distinction for Eye Movement Biometric Performance

Prior research states that sine-wave frequencies below 100 Hz carry the eye movement signal, and frequencies above 100 Hz can be considered noise. Here, we explore the biometric implications of this signal/noise distinction. We expect that there are important individual differences in the way subjects move their eyes, and this should lead to reliable biometric performance in the signal part. Although there is minimal eye-movement information in the noise part of the recordings, there may be important individual differences in the noise. The results suggest that the signal part contains the most important amount of identity-specific information, as anticipated. Nevertheless, the noise part performs substantially better than chance, and therefore must contain individual-specific information that distinguishes between individuals. This pattern holds for both short- (approximately 20 min) and long-term (approximately 1 year) biometric evaluations.