Combining online and offline handwriting recognition

Abstract

This work describes an online handwriting recognition system working in combination with an offline recognizer. The online input data is first transcribed by the online recognizer, then converted into an offline bitmap and recognized by the offline system. The outputs of the two recognizers are then combined probabilistically resulting in a classifier out-performing both individual systems. Experiments were performed over a database of single digits. The error rate of the online recognizer is reduced by 43% when the combination with the offline system is applied.