A Feature Fusion Based Approach for Handwritten Bangla Character Recognition Using Extreme Learning Machine

摘要

Optical Character Recognition (OCR) is an abstruse field of pattern recognition. An active branch of OCR is handwritten character recognition. This paper presents Bangla handwritten character recognition based on a feature fusion endeavor. Character recognition mostly depends on impeccable features extracted from input images. Coupling of two distinct feature vectors obtained by Histogram of Oriented Gradients (HOG) and Gabor filter is illustrated here. To evaluate the recognition rate of input characters Extreme Learning Machine (ELM) is used which is a feed-forward neural network. A 5-fold cross-validation scheme has been applied to measure the fulfillment of the organization. While using individual feature extraction technique, HOG and Gabor filter show 90.5% and 91.2% accuracy respectively. However, using feature fusion approach provides a better accuracy of 96.1%.