Hybrid Classifier-Based for Offline HandwrittenKannada Digit Recognition

The field of pattern recognition has many applications, and handwritten digit recognition is one of them. Sorting postal mail, processing bank checks, data entry forms, and other tasks are just a few examples of how handwritten digit recognition is used. This displays the data in a digitized format. We are releasing a new handwritten digit dataset for the Kannada script called Kannada MNIST (Modified National Institute of Standards and Technology), which may be used to directly replace the original MNIST dataset. This is made up of the digits 0 through 9. The appropriate parameters partition the Kannada MNIST dataset into training and testing. For Handwritten Digit Recognition, there are primarily two steps: feature extraction and digit recognition (HDR). The primary base for digit recognition is a set of categorization algorithms. Convolutional neural networks (CNNs) were used as feature extractors in the ongoing study. On the simple MNIST dataset, CNN is implemented using the Deep Learning Python framework, which provides an accuracy of 99.6%.We are adding a couple of extra classifiers to the CNN output to see which approach best supports the CNN Model for these digits. Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), and XG Boost are some of the classifiers we used. The output of the CNN is independently added with each of these classifiers. The output of the CNN model is feature extraction, which is fed to these classifiers to improve prediction accuracy. The major goal of this work is to develop a higher accuracy classifier by combining CNN and other classifiers.