Hi-index | 0.00 |
Hidden Markov Model (HMM) has been applied to the problem of machine recognition of Chinese handwriting. The character image is segmented into a number of local regions and feature vectors of these regions are extracted. The feature vectors are then used to get the observations for the HMM. The states of the HMM are to reflect the characteristic space structures of the character and its identities are obtained through the training samples usingsome algorithms. Two kinds of HMM are built and two more simple nearest neighbor classifiers (NN) based on the vector quantification process in the discrete HMM are employed. The combination of the classifiers is presented. Five kinds of features used to get the observations have been tried and three algorithms are adopted to determine the training process. The experimental result indicates the promising prospect of this approach.