This paper presents a Hidden Markov Model-based recognition system for Yoruba handwritten words (HMM). Data acquisition, preprocessing, feature extraction, and classification are the four stages of the work. Data were collected from adult indigenous writers, and the scanned images were preprocessed in various ways, including greyscale, binarization, noise removal, and normalization. The character shape, underdot, and diacritic were extracted from each of the normalized words, yielding a set of new features for handwritten Yoruba words based on a discrete cosine transform approach, and zigzag scanning was used to extract the character shape, underdot, and diacritic. sign derived from the spatial frequency of the word image The Yorb handwritten words were preprocessed to improve their quality, and the Discrete Cosine Transform was used to extract the features of the Yorub a handwritten image. The Yorub a words were modelled using a ten(10) state left-to-right HMM. Based on the model developed for the Yorub an alphabet, the initial probability of HMM was generated at random. In the HMM modeling, one HMM is used for each class of image feature. had been built For the handwritten word images, the Baum-Welch re-estimation algorithm was used to train each HMM class based on the DCT feature vector. The Viterbi algorithm was used to classify the handwritten words, yielding the corresponding state sequences that best describe the model. Our experiments yielded the highest test accuracy of 92 percent and the highest recognition rate of 95.6 percent, indicating that the recognition system's performance is very accurate.
Author (S) Details
Dr. Jumoke F. Ajao
Department of Computer Science, Kwara State University, Malete, Nigeria.
Stephen O. Olabiyisi
Department of Computer Science & Engineering, Ladoke Akintola University of Technology, Ogbomosho, Nigeria.
Prof. Elijah O. Omidiora
Department of Computer Science & Engineering, Ladoke Akintola University of Technology, Ogbomosho, Nigeria.
Oladayo O. Okediran
Department of Computer Science & Engineering, Ladoke Akintola University of Technology, Ogbomosho, Nigeria
View Book :- https://stm.bookpi.org/CTMCS-V2/article/view/1777
No comments:
Post a Comment