On-line Script Recognition

Automatic identification of handwritten script facilitates many important applications such as automatic transcription of multi-lingual documents and search for documents on the Internet containing a particular script. The increase in usage of handheld devices which accept handwritten input is creating a huge volume of handwritten data. We propose a method to classify words and lines in an online handwritten document into Arabic, Cyrillic, Devnagari, Han, Hebrew and Roman scripts. The proposed classification system, based on spatial and temporal features of the strokes, attained an overall classification accuracy of 86.5% at the word level on a dataset containing 13, 379 words. The classification accuracy improves to 95% as the number of words in the test sample is increased to five and to 95.1% for complete text lines.