English Handwritten Character Recognition Based on Ensembled Machine Learning

In recent days there are many advancements in optical character recognition (OCR), still, handwritten character recognition remains a challenge due to practices of realizing characters in many ambiguous forms. Currently, multiple algorithms based on deep learning can recognize a character in different languages like English, Devanagari, Chinese, etc. Existing methods have claimed to have an accuracy rate of up to 99 % . However, this accuracy is justified only for documents that are printed with fine text, but for degraded image data, these algorithms could not translate handwritten text into a recognized text with satisfactory performance. This work presents a state-of-the-art Novel Naive Propagation (NNP) Classification algorithm along with Harmonized Independent Component Analysis (HICA) and Hybrid Fireflies-Particle Swarm Optimization(HFPSO), which are used to extracting and selecting features from the image data, respectively. Due to the complexity of handwritten characters, the process of character recognition remains challenging. So, we have experimented with an ensembled classifier that combines the various components of the Naive Bayes Propagation Classification algorithm along with the Feed-forward and Backpropagation Neural Network. The experimental results and its analysis with various strategies show the better performance of the proposed system as compared to other techniques. Based on our experimentation we have identified that compared to other character recognition approaches, the Novel Naive Propagation Classifier is more advantageous for creating an automatic HCR system.