[Artifact detection and signal reconstruction in the pupillogram]

Just as the EEG reflects different wake and sleep stages, changes in pupil dynamics reflecting different levels of vigilance are also to be found. The literature contains numerous reports on experimental set-ups for the recording of the pupillogram. Interesting methods of signal processing are to be found in [7] and [5]. Currently, such recordings are being used to check the success of sleep therapy. A problem that still needs solving is the optimal handling of artifacts caused by blinking. The present article proposes a procedure for artifact detection by back-propagation networks, and subsequent reconstruction of the signal by an AR model. Estimation of the signal is first demonstrated by a test signal, and then by a corrupted pupillogram.