Sign language teaching protocol with an interactive robotic arm and real-time vision feedback

BACKGROUND: Sign Language (SL) is essential for Deaf communities, yet limited SL proficiency sustains communication barriers in education and services. We propose an interactive tutoring system for Ecuadorian Sign Language (EcSL) that combines real-time, vision-based gesture assessment with an embodied "reward" delivered by a robotic arm, using fruit vocabulary as an initial learning domain. METHODS: We will conduct a two-arm randomized controlled experimental design, pretest-posttest control-group study with individual random assignment. About 70 university students without formal EcSL training will learn EcSL fruit signs through video and sign recognition feedback module. The system uses a webcam, MediaPipe landmark extraction, and a landmark-based fusion classifier that combines a local bidirectional LSTM with a transferred ST-GCN motion encoder; modules communicate via Robot Operating System (ROS). Outcomes. A total of six dependent variables will be assessed in both the experimental and control groups: (a) handshape, (b) palm orientation, (c) location, (d) movement, (e) non-manual signals and expressions, and (f) global coordination. Analyses will use descriptive statistics, reliability checks when applicable, and a mixed repeated-measures MANOVA. DISCUSSION: By coupling immediate feedback with embodied reinforcement, delivered through a robotic arm, the platform aims to support early sign accuracy and provide a motivating practice environment for learners. The robotic arm responds to participants' performance with physical rewards, thereby reinforcing correct execution through tangible, real-time stimuli. Results will guide improvements in robustness and inform extensions to broader vocabularies and deployment in educational contexts.