A Novel Empowered System for ISL Gesture Recognition

Abstract

In India, a large population is of hearing-impaired and vocal disabled. To provide them equality in society and to reduce their isolation in society an Indian Sign Language (ISL) for gesture recognition is highly required. This paper introduces an improved novel empowered system for ISL gesture recognition which can recognize accurately live videos including finger word spelling, and single-handed static and dynamic gestures. In the proposed system, Artificial Intelligence (AI), and Machine Learning (ML) approaches are incorporated to promptly identify the acknowledge and interpret ISL gestures. It also removes the communication barriers and fostering inclusiveness to improve the accuracy of the proposed system. Additionally, to improve the accuracy the Convolutional Neural Networks (CNNs) approach is incorporated to extract hierarchical features from visual data, enabling precise and prompt sign language interpretation, and the Support Vector Machines (SVMs) approach is incorporated to optimize classification margins and handles high-dimensional data effectively. With a 91% accuracy rate for single-handed dynamic words and a 97% accuracy rate for finger spelling alphabets, the proposed method successfully detects both types of words. Experiment outcomes indicate that this method attains a higher recognition rate than various existing techniques.