Sign Language to Voice and Display Conversion Using Deep Learning on Raspberry-Pi
Abstract
This paper presents a deep learning-based sign language recognition system designed to improve communication accessibility for people who have significant hearing loss or who experience challenges in hearing. The system employs computer vision and convolutional neural networks (CNNs) to classify hand gestures representing sign language letters. Image preprocessing techniques such as resizing and normalization are used to improve recognition accuracy. The CNN architecture incorporates ReLU activation and max-pooling layers for efficient feature extraction, followed by a softmax layer for multi-class classification. Real-time gesture tracking using OpenCV ensures temporal awareness by analyzing finger joint positions. Recognized gestures are converted into text and further synthesized into audible speech using advanced text-to-speech (TTS) technology. A Tkinter-based graphical user interface (GUI) displays recognized signs and offers context-aware text prediction. Experimental results demonstrate Accurate and reliable real-time translation of sign language into text and speech.