Real-Time Sign Language Recognition System: A Computer Vision-Based Approach for Accessible Human–Computer Interaction

This paper, titled “Real-Time Sign Language Recognition System: A Computer Vision-Based Approach for Accessible Human–Computer Interaction,” proposes a camera-based system for recognizing hand gestures and converting supported signs into readable text. The proposed system uses a modular processing pipeline consisting of video acquisition, hand detection, landmark and feature extraction, gesture classification, temporal smoothing, and text output. Python, OpenCV, MediaPipe, and a machine-learning classifier are suggested as potential implementation technologies. The system is designed to capture live video frames, identify hand landmarks, normalize coordinates, extract relevant features, classify supported gestures, and display the predicted sign. Confidence thresholds and temporal smoothing can help reduce unreliable or inconsistent predictions. The design also supports future extensions such as speech output, dynamic gesture recognition, and expanded sign vocabularies. The paper describes the proposed architecture, system modules, processing algorithm, implementation approach, evaluation plan, limitations, and future scope. Potential evaluation measures include classification accuracy, per-class precision and recall, inference latency, and processing frame rate. This work presents a proposed software-system design intended to improve communication accessibility through computer vision and machine learning. Experimental performance results are not claimed in the current paper; implementation and testing are required to validate the system's effectiveness.