Sep. 17 2007 2:00PM

Sep. 17 2007 3:00PM

This talk presents a new model in which the cassifier receives not only a pattern itself but also supplementary information that assists recognition. This model enables us to achieve a 100%
recognition rate with a 0% rejection rate with certain bits of supplementary information required. For printed characters, experiments show that 4 bits of supplementary information were
required in the leave-one-out method and 1 bit was in the resubstitution method. In addition, we generalize the discussion into the relationship among a quantity of supplementary information,
a recognition rate and rejection rate. The theory presented in this talk is applied to the data embedding of a font set for camera-based character recognition.