Hand Written Character Feature Extraction Using Non-Linear Feedforward Neural Networks

• DOI: 10.6688/jise.2005.21.2.10
• Publication Date: 2005-03-01
• Authors: Abdul Jalil; Ijaz Mansoor Qureshi; Tanweer Ahmad Cheema; Aqdas Naveed
• Keywords: Feedforward neural network; Competitive learning; Feature (linguistics); Feature vector

In this paper, an artificial neural network is proposed for feature extraction of hand written characters. The learning algorithm developed based on a modified Sammon's stress our feedforward networks, which can not only minimize intra class pattern distances but also preserve interclass in the output space. method tries to calculate rough classes using Competitive Learning network, unsupervised network. Then was used with perform information obtained by means Network. features thus were compared standard PCA and terms their classification accuracy. Two numerical criteria performance evaluation features-the normalized error rate stress. It found that provides are more efficient these two criteria.