Feature Extraction and Dimensionality Reduction in Pattern Recognition Using Handwritten Odia Numerals
Pradeepta Kumar Sarangi, Kiran Kumar Ravulakollu · 2014
Feature extraction is the initial and critical stage which needs to be carried out very carefully for any recognition system that uses pattern matching. In order to reduce the feature extraction complexity, dimensionality reduction is applied. This also increases the performance and recognition accuracy. This paper proposes the concept of a new feature extraction and dimensionality reduction method based on a set of linear transformation of the character image. The verification of the method has been carried out by implementing a simple recurrent neural network (RNN) with a data set consisting of 1500 isolated handwritten Odia numerals. An accuracy of 92.41% is reported. Experimental results show that the proposed method has the potential to be used as a feature extraction method for handwritten Odia numerals.