A Visual Acuity Assessment System Based on Static Gesture Recognition and Naive Bayes Classifier

Abstract

Visual Acuity (VA) assessment is crucial for early vision screening, yet traditional methods are manual and time-consuming. Despite the advancements in human-computer interaction (HCI), there is no existing system fully addresses accuracy, efficiency and adaptability. This study introduces an intelligent VA assessment system that combines MediaPipe-based static gesture recognition with a novel Naive Bayes Classifier (NBC)-based VA Thresholds Determination (VATD) scheme. This integration offers a non-contact, user-friendly approach for rapid and precise VA testing. The VATD scheme is designed to significantly reduce the number of test trials; thereby, substantially improving the efficiency. Experimental validation confirms the system's high accuracy (96.72%) within a ±0.1 deviation from standard methods, achieves a 68% reduction in test time compared to traditional methods, and offers a 27% efficiency improvement over ANN-based systems. This system promises to enhance VA assessments, particularly for children and adolescents, with its speed, accuracy, and broader applicability.