ABSTRACT
Recently, Deep Convolutional Neural Networks (CNNs) have become one of the most appealing approaches and they have been a crucial factor in the variety of recent successes and challenging machine learning applications such as object detection and face recognition. In the current age of digitalization, handwriting recognition plays an important role in information processing. The handwriting recognition system aims to convert handwritten digits into machine-readable formats. Handwriting digit recognition application is used in different tasks for real-life time purposes such as vehicle license-plate recognition, banks for reading checks etc. All these areas aim to deal with huge databases and so demand high recognition accuracy, lesser computational complexity, and consistent performance of the recognition system. In this research work, a convolutional neural network (CNN)-based digit recognition technique has been proposed. The ability to automatically detect the important features of an object (here an object can be an image, a handwritten character, a face, etc.) without any human supervision or intervention makes CNNs more efficient. Optical Character Recognition (OCR) is a subfield of Image Processing that is concerned with extracting text from images or scanned documents. The work of OCR is to recognize handwritten characters. In this chapter, a mechanism that will recognize handwritten digits of different languages has been proposed. Here MNIST dataset, CMATERDB dataset, CMATERDB/Devanagari dataset, and CMATERdb/Telugu dataset have been used to recognize digits. A CNN framework has added pooling layers into the CNN layers. The model also focuses on the removal of discretionary details in the images of the digits and aims to enhance the overall characteristics. The ReLU function has been used as the activation function for the hidden layers of the network and softmax has been used as an activation function mainly for the output layer of the network. In this research work, the Adam optimization technique has been used in the algorithm in order to achieve better results in terms of both accuracy and performance.
