A Hybrid Deep Learning Framework for Handwriting Language Detection Using CNN and SVM

The multilingual handwriting identification system uses a hybrid model of convolutional neural network(CNN) and support vector machine (SVM), analyzing a real-time 10,000-image dataset in multiple languages, including English, Hindi, Punjabi, Urdu, and Gujarati. A total of two thousand photos per class have been collected for classification. This work uses a CNN for feature extraction, while an SVM is used for classification. The suggested approach to classifying the image using SVM includes three convolutional layers, three max pool layers, one floating layer, and a fully connected layer. Fully linked layers are used for classification to reduce dimensionality through layer pooling. The current study employed the preprocessed dataset to train the CNN hierarchical features to various scripts and languages. With a precision of detection of classification based on language detection has been done with an accuracy of 95.48%. The model’s primary goal is to recognize handwriting images in the digital realm. The model’s main contribution is its ability to properly identify and digitize handwritten text in many languages, making document management, translation, and archiving more efficient. Furthermore, developments in this area foster the creation of increasingly complex artificial intelligence systems and spur creativity in applications such as computer vision and natural language processing.