Moment Based Online and Offline Handwritten Character Recognition

Automating communication between human and machine i.e. interpreting human writing provides huge benefits in automatic processing of bulk amount of papers, transferring data into machines and web interfaces to paper documents. There has been intensive research in progress on various scripts such as English, Japanese, Chinese, and Indian. In this paper we present a system that recognizes handwritten characters of Devanagari, the most popular script in India. Handwritten character is inputted using digitizing pen tablet to capture the sequence of (x,y) coordinates. Based on online data we have calculated 1-D geometric moments, 1-D Central moments, 1-D Hu's moments and 1-D complex moments and using offline data we have calculated 2-D geometric moments, 2-D Central moments, 2-D Hu's moments and 2-D complex moments. MLP-BP Neural Network is used as Classifier for classification. Features extracted using moments and class Input is fed to Classifier. The classifier derives the knowledge base which will be used as a basis for testing to recognize the handwritten character.
We calculated recognition result of individual moment features as well as its combined effect also analyzed. We got 74% recognition result for offline central moment and 71% for online central moment when we considered them as individual. We also observed that after combining features we got improved recognition result.