Desi Fatkhi Azizah, Anik Nur Handayani, Aji Prasetya Wibawa
Objectives Communication barriers experienced by people who are deaf remain a persistent public health challenge, particularly in healthcare settings where effective interaction is essential for service quality, patient safety, and equity. This study aimed to evaluate an artificial intelligence (AI)–assisted sign language recognition system as a pre-implementation digital health intervention to improve access to health communication for people with hearing loss. Study design Pre-implementation quantitative system evaluation. Methods This study employed a quantitative evaluative design to assess the feasibility of an AI-assisted sign language recognition system using a British Sign Language (BSL) alphabet image dataset. System performance was evaluated across four data partitioning scenarios using multiple model configurations. Performance metrics, including precision, recall, F1-score, mean Average Precision (mAP), and inference time, were interpreted as indicators of feasibility, reliability, and responsiveness relevant to health communication contexts. Descriptive analysis was conducted to examine performance consistency across scenarios. Results Across all evaluation scenarios, the system demonstrated consistently high performance and stable responsiveness, indicating reliable recognition of sign language gestures under varying data configurations. Performance consistency across scenarios suggests that system feasibility was not dependent on a single data split configuration. All model variants achieved performance levels considered adequate for supporting basic sign language–based communication, with expected trade-offs between recognition accuracy and response time. Conclusions From a public health perspective, the findings indicate that AI-assisted sign language recognition systems are feasible as complementary digital health communication tools, particularly in settings with limited access to professional interpreters. While this study does not assess clinical outcomes or real-world implementation, it provides early-stage evidence to inform future implementation-based evaluations and policy considerations aimed at promoting more inclusive and equitable healthcare services for deaf people. Copyright © 2026. Published by Elsevier Ltd.
Department of Electrical Engineering and Informatics, Faculty of Engineering, Universitas Negeri Malang, Semarang Street 5, East Java, Malang, 65145, Indonesia
