UNIVERSAL BIDIRECTIONAL SIGN LANGUAGE TRANSLATION USING DEEP LEARNING

This paper presents the design and implementation of a sign language translation system focused primarily on converting text or speech into sign language, while optionally supporting the reverse sign to text/speech.The proposed system harnesses the power of deep learning techniques in conjunction with fuzzy logic to address the ambiguity and variability inherent in sign languages.To manage the sequencing of gestures and contextual cues, Recurrent Neural Networks (RNNs) or Transformer-based architectures are integrated into the pipeline.These deep learning models are further enhanced through the inclusion of a fuzzy logic layer, which interprets soft boundaries and uncertainty in gesture rendering, ensuring more natural and adaptable animation outputs.The core functionality translates text or spoken commands into sign language through animated avatars /prerecorder video clip / 2d pose animation, allowing for real-time communication from non-sign language users to the signing community.In the optional reverse direction, sign language input can be captured and converted into text or synthesized speech, providing flexibility for expanded use cases.Deployment on mobile platforms is enabled through lightweight model optimization techniques (e.g., TensorFlow Lite or PyTorch Mobile), ensuring low latency and responsiveness.This framework is built to support multiple sign language datasets, including ASL, BSL, and other international sign languages, promoting inclusive and accessible communication across diverse linguistic communities.