Recognition of Typewritten Characters Using Hidden Markov Models

This paper presents a typewritten characters recognition system using Hidden Markov Model (HMM). Character recognition systems convert images of printed, typewritten or handwritten documents into computer readable texts that can be easily edited or searched. Character recognition for typewritten documents is however difficult due to broken edges, touching characters, shape variance, skewing, and heavy printing resulting from the typewriter impact. Three documents (old memo, old war letter and newly typewritten essay) were used to create three datasets of typewritten characters each consisting of 1995, 702 and 2049 characters respectively. The research result showed that, recognition accuracy values are 94.88%, 91.45% and 97.24% for old memo, old war letter and newly typewritten essay datasets respectively. Hence, HMM is an efficient method that can be employed to recognise typewritten documents.