Artefact 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 recor ding of the pupillogram. Interesting methods of signal processing are to be found in [7] and [5]. Currently, such recordings are being used to check t he success of sleep therapy. A problem that still needs solving is the optimal handling of artifacts cau sed by blinking. The present article proposes a procedure for artifact dete ction by backpropagation networks, and subsequent reconstruction of the sig nal by an AR model. Estimation of the signal is first demonstrated by a tes t signal, and then by a corrupted pupillogram.