Geometrical, profile and HOG feature based recognition of Meetei Mayek characters
Chandan Jyoti Kumar, Sanjib Kr. Kalita · International Conference on Computing for Sustainable Global Development · 2016
Recognizaton of Handwritten characters is a challenging task in the computer vision field. However, if can be done with good accuracy will be very much useful for digitizing various documents. Researchers have made various investigations for developing optical character recognition system. The accuracy of the system highly depends upon the features and classifier used. A lot of work is carried out at International level, in India also lot of work is reported in major regional scripts; Devanagri, Bangla, Gurumukhi are few of them. However, it has been observed that only few researchers have considered working on North East Indian regional scripts. As the scripts vary from one another a lot, the feature working well for one script may not be suitable for another one. Our main focus is to develop an OCR system which can work well for Meetei Mayek script. For recognition of handwritten Meetei Mayek isolated characters, we have considered Geometrical and Histogram oriented gradient feature and profile features, which are used as input to the classifier. Recognition accuracy is calculated for SVM and a comparative study is carried out.