Geometric Features and Template Matching for English Sentence Handwritten Identification System

Tanasanee Phienthrakul
This research presents a handwriting identification system, developed by geometric features and template matching features on Android platform. To recognize the handwriting in English language, the proposed system uses English sentences as the training data. For geometric features, average height/width of each character, average ratio, average angle, and space between characters and words are taken into the consideration. To increase the system performance, a template matching is used to extract more features by consideration in 5 groups of simple pattern which are looping, crossing, branching, curving, and turning points. After applying the feature extraction, we use this data model for classification to verify the writers by k-nearest neighbor. In the application, users are able to choose an option whether to take a new input picture (picture of handwritten document) or to select a picture from the phone gallery. The particular picture selected by users will be sent to the server for obtaining the result. Then, the result will be fed back to the mobile application on screen. The experimental results show that the proposed technique can identify the handwriting image with high accuracy rate.
Handwriting Identification; Feature Extraction; Template Matching; Geometric Features; k-Nearest Neighbor