HMM Based On-line Handwritten Chinese Character Recognition

How to use computer to process and recognize words is an active research area in pattern recognition. With the popularity of mobile devices, the traditional method, inputting Chinese characters via keyboard, is not satisfied with the requirements of people. To address this problem, on-line handwritten Chinese character recognition (OLCCR) has been paid much attention in past several decades due to its more direct, flexible, convenient and acceptable characteristics. In this dissertation, preprocessing, coarse classification and feature extraction in Hidden Markov Model (HMM)-based OLCCR is studied to recognize not only regular written characters but also characters with stoke-connected. The main contents of dissertation are as follows:1. The dissertation reviews the development of on-line handwritten Chinese character recognition and analyzes the on-line handwritten Chinese character in details. Based on this content, the scheme for on-line handwritten Chinese character recognition is proposed by using HMM, which efficiently uses the structural and statistical features of the hand-written Chinese characters.2. A novel segment extraction method is proposed to address the problem of connected-stroke characters. The proposed method partitions each character into segment sequences, then three attributes in characterizing strokes, real, virtual and null segments, are assigned to each segment in a character. According to the decision rules, a connected-stroke style character is transformed into a regular style character. Experiments show the high efficiency of the proposed method.3. Chinese character recognition is a typically large pattern recognition. Coarse classification is necessary to reduce the number of candidate characters before fine matching. The coarse classification achieved good result using long segment, stroke number and structure type.4. The recognition program is transplanted to a Samsung SGH-i718 smartphone, which contains the Windows Mobile 5.0 operating system. Then, embedded handwritten Chinese character recognition software is realized in the smart phone.