Implementasi Teknik Gamma Correction pada Sistem Penerjemah Bahasa Isyarat

Abstract

It’s known that approximately 1.3 billion people or 16% of the global population experienced significant disabilities in 2023. Individuals with impaired hearing and speech use sign language as a communication tool. One solution to bridge communication between disabled people and the general public is a sign language translation system capable of detecting hand movements and translating them into text. A major challenge in detecting hand movements is insufficient lighting, which can affect system performance. This study employs Gamma Correction as an image enhancement technique to improve image quality by adjusting brightness, allowing hand movements to be more easily detected. Gamma Correction is applied to a webbased system using the OpenCV.js library, which modifies every pixel value in images. The dataset in this research consists of 10,000 hand gesture images divided into 25 classes, in which 7,500 images are used as training data and 2,500 images as validation data. The testing was conducted by comparing a Convolutional Neural Network model's performance on validation data without Gamma Correction and with Gamma Correction. The results of testing show that even though there’s no significant increase in training accuracy, the model's generalization ability improved. Additionally, analysis through the Confusion Matrix revealed a 2% increase in accuracy. This model will be integrated into a website to translate hand movements into words or sentences.