Real-Time Indian Sign Language Recognition Using CNNs for Communication Accessibility

The challenge of communication for the deaf and mute community continues to pose a barrier in connecting with society. Sign language, a manual communication method, has emerged as an essential tool for this group, yet it remains largely unrecognized by the majority of the population. This research proposes a machine learning-based Indian Sign Language (ISL) detection system utilizing Convolutional Neural Networks (CNN) to bridge this gap. The system is designed to automatically recognize hand gestures representing ISL alphabets in real-time through a camera interface. Key steps include image preprocessing, gesture detection, and classification using a trained CNN model, followed by deployment on mobile platforms via TensorFlow Lite integrated with Flutter. This approach ensures the model is lightweight yet capable of delivering high accuracy in real-world settings. The model achieves impressive results, with accuracy levels exceeding 90% in predicting hand gestures. The application is user-friendly, enabling anyone with a smartphone to recognize ISL symbols and assist in communication with the deaf-mute community. This paper discusses the implementation, performance, and potential extensions of the system, positioning it as an effective tool for improving communication accessibility.