Bangla Online Handwriting Recognition System

C-DAC Logo

Handwritten character is the aggregation of sequence of strokes. In our approach, an on-line handwritten character is characterized by structure- or- shape-based representation of a stroke in which a stroke is represented as a string of shape features. Using this string representation, an unknown stroke is identified by comparing it with a database of strokes using elastic matching technique On-line handwriting data are a dynamic, digitized representation of a (digital) pen movement, generally describing sequential information about position, velocity, acceleration or even pen angles as a function of time. On line handwriting recognition takes on a novel significance in the context of Indian languages. At present word processing in Indian languages can be a troublesome experience, considering the restriction on use of the regular keyboard, designed for English. Keyboard mapping systems are used in case of Indian languages, which are not convenient to use. A comfortable solution would be to let the user write in a natural, normal fashion using a suitable pen-like device and let the computer do the rest. That transfers the burden of learning keyboard mappings from the user to the computer.

Our approach:

Following are the steps for our Bangla OHR system:

  1. Pen Input
  2. Preprocessing
  3. Segmentation
  4. Feature extraction
  5. Stroke Identification
  6. Character recognition.
The alphabet of modern Bangla script consists of 11 vowels and 39 consonants, known as basic characters. Writing style in the script is from left to right. The concept of upper/lower case is absent in Bangla script. A recognition experiment has been conducted with a total of 495 classes, which includes compound characters, symbols, ligatures etc. used in Bangla handwritten text.

Data collection is done using Pen Tablet. The front-end for data collection is developed using Microsoft Tablet PC SDK. Preprocessing consists of normalization and smoothing of strokes. We have identified some critical points and essential shape features, which represents strokes for Bangla character. Finally elastic matching classifier (Dynamic Time Warping) does the necessary job for character classification. The same technique can be used for recognizing other Indian languages.

Fig.1. Flow Diagram of Bangla OHR

System Features :

Software Requirements :

Hardware Requirements :

Intel Pentium 4 or AMD Athlon, 64 bit/32 bit processor (2GHZ or faster), 100 MB of available hard disk space for installation: additional free space required during installation. Minimum 1 GB of RAM (2 GB recommended).