Machine Learning and Vision Based Techniques for Detecting and Recognizing Indian Sign Language

Abstract

Despite rapid global advancement in technology, the persistent challenge of hearing impairments affects a significant proportion of the global population. Individuals with these impairments face complex communication barriers daily. The Indian Sign Language (ISL) has emerged as a universal communication tool for individuals with hearing impairments in India, playing a vital role in educational institutions and bridging societal gaps in a country marked by rich cultural and linguistic diversity. This work presents an innovative Supervised Learning approach for ISL recognition that extends beyond traditional classification techniques. The method employs advanced algorithms designed to classify new observations. Utilizing an expansive dataset, intricate patterns and nuances are identified, fostering accurate and adaptive decision-making. A Convolutional Neural Network (CNN) algorithm is applied, not only for data classification but also for iterative learning and refinement of classification boundaries. In the vast expanse of n-dimensional space, the CNN strives to identify optimal hyperplanes, establishing dynamic decision boundaries to adeptly categorize diverse data points. This approach transcends traditional classification boundaries, offering a more nuanced and effective data-driven decision-making process. This research heralds a new direction in addressing communication barriers, with potential applications extending beyond the realm of ISL. The assimilation of numerous regional languages into the sign language matrix, whilst challenging, is key to fostering a life of normalcy and integration.