Handwritten Pattern Recognition System Using Artificial Neural Network

Abstract

Artificial Intelligence is proving to be vital in the 21st century. Difficult tasks are being performed with the help of many applications. But the areas like computer vision lacks accuracy to recognize patterns and letters. The accuracy to recognize handwritten pattern is still less than the expectation. Handwritten digit recognition has been a major problem in the computer vision. Present algorithms are inaccurate to make it a perfect smart system. This is an important field because of many day-to-day life applications where large number of documents with handwritten digits should be entered and converted into text form such as bank cheque analysis, insurance paper analysis and other handwritten forms etc. So, our main aim is to develop a system program which can tackle with handwritten digits and letters and makes it easy to read by the computer system. We are going to use Machine Learning algorithm such as Backpropagation Neural Network Algorithm to build a system which can deal with the problem of computer vision with higher accuracy. Additionally, we are going to implement Multiplier Feed Formal Neural Network and Support Vector Machine algorithms to recognize various other patterns. In order to achieve 98% accuracy, we are going to focus on different Image processing techniques also. Large number of handwritten datasets will be used to train our system. After that, our trained system will be able to understand unknown complex handwritten patterns.