International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 05 | May 2024
www.irjet.net
p-ISSN: 2395-0072
Survey Paper on Hand Written Digit Classifier Anushka Singh Chauhan1, Aryan Srivastava2, Er. Sandeep Kr. Dubey3, Er. Anuj Singh4 1UG student of Department of Computer Science and Engineering , Shri Ramswaroop Memorial College of
Engineering and Management Lucknow, Uttar Pradesh, India
2UG student of Department of Computer Science and Engineering, Shri Ramswaroop Memorial College of
Engineering and Management Lucknow, Uttar Pradesh, India
3Associate Professor, Department of Computer Science and Engineering, Shri Ramswaroop Memorial College of
Engineering and Management Lucknow, Uttar Pradesh, India
4Associate Professor, Department of Computer Science and Engineering, Shri Ramswaroop Memorial College
of Engineering and Management Lucknow, Uttar Pradesh, India -------------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Recognition of the handwritten digits is
a basic question in the field of computer vision and machine learning. The review of the current literature gives an understanding of techniques and algorithms involved in recognizing hand-written digits. The scope of the study involves a number of techniques such as convolutional neural networks, deep learning, and what has to do with the neural network algorithms. There are various kinds of classifiers which are used for training these algorithms and testing them, and this is facilitated by the availability of various datasets including the MNIST database. It also discusses about the newer approaches like Bayesian networks and Turbo decoding for better recognition and stability of the hand-written digit methods. Moreover, the survey also elaborates on the technique called orthogonal feature detectors and the generation if binary templates required for the faster classification. In conclusion, this paper offers a review on the state of the art on the subject of handwritten digit recognition hence establishing benchmarks on the subject as well as pointing out areas that need improvement.
1. INTRODUCTION Handwritten digit recognition is one of the easiest and fundamental problems in the context of machine learning and artificial applications. Investigations on the determination of the most effective approach to digit recognition have been conducted with uses of methods that subsumes advanced architectures including deep learning. algorithms like the shallow feedforward neural networks, convolutional neural networks, and the support vector machines have been used to come up with higher classification accuracies. These methods have been used and tested on set of data such as the MNIST database that contains a set of hand written digits mainly from “0” to “9”. This relatively new methodology of classification of hand written digit has enable investigation of some recent techniques that are the use of Bayesian networks for decoding-classification recovery and the use of the connectivity information
© 2024, IRJET
|
Impact Factor value: 8.226
|
from fMRI data in dealing with hand written digit. It is also important to note that there are successful attempts to improve the performance of the recognition process using deep learning techniques. It also explored the ways to enhance the efficiency, stability, and accuracy of the outcome for classification in reference to various forms of hand-written characters. By performing the classification of hand-written digits it not only extends the knowledge base of artificial intelligence but can also be applied to other technologies such as Optical Character Recognition (OCR) systems, and other research in the field of pattern classification. Comparing and analysing various methodologies and envisaging better algorithms in the field the researchers attempt at proposing better hand written digit classifiers to keep the scope of categorisation higher and recognition systems better in other domains as well.
2. LITERATURE SURVEY The underlying task of the proposed sources is the problem of the automatic interpretation of intelligible handwritten input, which is of interest for the pattern recognition research community since it can be applied in many fields to replace existing input devices that require more effort to organize and process data by means of more easily usable ones. Generally, handwritten digit recognition can be discussed one of the fundament issues when developing perspective recognition systems. Some of the areas that can benefit from digit recognition techniques include post office automation, automatic reading of addresses and subsequent sorting and routing of mails, check cheque processing among others. [Error! Reference source not found.] In reviewing the various approaches and methodologies applied in the field of handwritten digit classifiers, one realizes that there are many different approaches and methodologies possible. Lecun et al. (1998), particularly extended the convectional gradient-based learning method and introduced document recognition steering with special attention to written digit recognition (Cireşan et al.,
ISO 9001:2008 Certified Journal
| Page 1503