A Neural Network Based Recognition of Isolated Handwritten Basic Bangla Alphabets and Numerals Using Spatial Relationships

Manoara Begum, Fujaila Shahnaj, Tahamida Aktar, Yeasmin Ara Akter, Nayan Kumar Nath

Abstract


Due to the different shapes, sizes, and handwriting styles, selection of proper feature set is more important to get high recognition accuracy. This study aimed to evaluate the performance of isolated handwritten basic characters and digits’ recognition using zoning accompanied by spatial relationships, especially directional relationships. This paper is based on diagonal element feature extraction approach for the recognition of handwritten Bangla characters. This system is implemented with samples of 10,000 handwritten Bangla characters and also with a database of 10,000 samples of Bangla numerals. Experimental result reveals that the success rate is 88.34% for Basic Bangla characters and 92.43% for Bangla numerals.

Aus. J. Bus. Sco S. & IT. Vol 4(1), January 2018, P 33-39


Keywords


Handwritten Characters; Adaptive Weighted Median Filter; Normalization; Spatial Relationships; Back Propagation

Full Text:

PDF

Refbacks

  • There are currently no refbacks.