Real-Time Indian Sign Language Recognition Using Hand Landmarks and ANN with Hindi Speech Output
Year: 2026
Type: conference-paper
Abstract: This proposed work aims to narrow the gap between those who know Indian Sign Language (ISL) and those who do not know it. The authors have introduced an AI-based system that can convert ISL gestures into Hindi text and speech. The idea is to recognize hand movements using a webcam and convert the action into text and speech in real-time. The system has used MediaPipe to recognize 21 landmarks on the hand and combine them into a 63-dimensional vector. After normalizing this vector, it has used an Artificial Neural Network (ANN) to classify the gestures into words. The data includes various ISL gesture classes under different lighting conditions to increase robustness. For converting text into speech, Google Text-toSpeech (gTTS), is used to converts the text into spoken language in real-time. The system has used TensorFlow and OpenCV to implement this system. The experimental results show an accuracy of around$94-95 {\%}$, and it has low latency, and suitable for real-time usage with typical hardware and webcam. The idea is to make ISL more accessible and communicative by converting it into Hindi text and speech.
Institution: Chandigarh University
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7167679136