A study presenting a lightweight machine learning model for Arabic Sign Language recognition using Mediapipe hand-landmark features, achieving high accuracy on both public and newly collected Lebanese-dialect gesture datasets while requiring far fewer parameters than YOLO-based methods.