The Language of Faces: From Gestures to Text

With the rapid growth of artificial intelligence and computer vision, technology is now playing a key role in bridging communication barriers for the deaf and hard-of-hearing community. Indian Sign Language (ISL) relies on a combination of hand gestures and subtle non-manual cues such as facial expressions, head orientation, and body posture to convey meaning and emotion. However, most existing translation systems focus only on hand movements, leading to partial or unclear interpretations. To address this limitation, the proposed system integrates Mediapipe Face Mesh for detecting facial landmarks along with Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models for visual feature extraction and sequential pattern recognition. By combining these techniques, the system efficiently converts both manual and non-manual features of ISL into accurate and context-aware text. This work enhances the reliability and naturalness of sign language translation, promoting better digital accessibility and inclusivity for individuals with hearing impairments.