Hybrid LSTM-Transformer Architecture with Generative AI for Enhanced Indian Sign Language Recognition

Sign language recognition plays a crucial role in bridging communication gaps between the deaf and hearing communities. Despite its importance, Indian Sign Language (ISL) remains underrepresented in technical research. This study introduces an advanced multimodal ISL recognition system leveraging a hybrid LSTM-Transformer architecture and generative AI for text-to-speech (TTS) synthesis. The model achieves over 97% accuracy across seven key gestures, addressing challenges such as dataset quality and environmental variability. By integrating generative TTS, the system enables natural, contextually relevant communication. This research contributes to the inclusivity and accessibility of ISL technology in India.