Handwritten character recognition by combining multiple recognizers using conditional probabilities

This research mainly focuses on recognizing the handwritten characters on a form in order to automate the Medical Form processing. Several efficient algorithms have been developed by us so far, to separate the handwritten characters from printed text character; to separate the lines, words and each character. In this thesis, we concentrate on the recognition of the segmented handwritten characters. Different feature recognition algorithms are employed and their performance on a given training set is analyzed. We find a way to combine all these individual feature recognition algorithms by incorporating their interdependence. The reliability of these algorithms is determined in terms of Conditional Probabilities and a rule for classifying the input character based on the outputs of each individual feature recognition algorithms is identified from the observations