The tutorial provides an overview of state-of-the-art hardware solutions for sign language data acquisition, including wearable sensors, vision-based systems, depth cameras, and inertial measurement units, with a focus on their metrological characteristics and limitations. From a methodological perspective, the tutorial reviews algorithmic approaches for sign language analysis, covering feature extraction, signal preprocessing, classification, and machine learning pipelines commonly adopted in AI-based recognition systems. Particular attention is devoted to the role of data quality, repeatability, and uncertainty, highlighting how metrological aspects directly influence the performance and reliability of AI models.
A hands-on session is included, where participants will work with pre-acquired datasets to explore supervised classification techniques using MATLAB’s Classification Learner toolbox. Through practical examples focused on the recognition of sign language alphabet letters, participants will gain insight into model training, validation, performance evaluation, and the critical impact of dataset quality before AI training.
