Indian Sign Language Detection Using Lightweight Architecture

The detection of Indian Sign Language (ISL) is highly essential to the deaf and typically hearing population in bridging the gaps in communication between the two groups. Detection of ISL through previous methods was shown to have more complex architectures, which often resulted in slower processing, thus posing a problem in real-world practical applications. To address this problem, a new method known as a few-shot prototypical network with a pre-trained DenseNet121 model was developed. Such an approach, designed to support only a few classes, has great classificatory accuracy and efficiency; indeed, satisfactory performance was proven to achieve 91 % accuracy, precision at 94.44 % and recall at 91.67 %. This verifies that the model is reliable and efficient. Simpler yet smarter detection methods can improve the speed and accuracy of ISL recognition. The enhancements to ISL detection represent a significant improvement in better communication among India's deaf community and make the technology more accessible for widespread use.