Gujarati language belongs to the Devanagari family; consists of characters and digits with typically sharp curves and few of them have similar shapes. Misrecognition of Gujarati Handwritten characters are mainly due to the existing similarities in their shape. To enhance reading capabilities of handwritten Gujarati characters and reducing optimal training time, automatic character recognition is the process in which the computers detect and identify the scanned/ captured characters. Major application area of Gujarati handwritten character recognition (HCR), includes post offices, schools, colleges, etc., in Gujarat. In this work a dataset has been prepared which consist of 170 Gujarati handwritten character (5 samples each of 34 Gujarati consonants), acquired with high resolution smartphone camera (13 MP 1/3.06 inch sensor). Further, an investigation on effectiveness of combination of grey level co-occurrence matrix (GLCM) and dynamic time warping (DTW) method for Gujarati HCR has been performed. The reason that GLCM features are used because it improves the recognition accuracy by reducing computation time and misrecognition of characters with similar shapes. Out of the three experiments performed, it has been observed that experiment 3 yields better recognition accuracy of 99.4% for the dataset under consideration.
展开▼
