Artificial Intelligence for interpreting static human arm signals
Milán Zsolt Bagladi · Az Eszterházy Károly Tanárképző Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Az Eszterházy Károly Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Annales mathematicae et informaticae · 2025
This paper presents a method for static arm signal recognition using OpenPose-based keypoint estimation, keypoint normalization, and two distinct classification approaches: K-means clustering and a neural network classifier. The system works with a simple camera setup and generalizes across users. A keypoint normalization technique is used to handle differences in body size and camera distance. To improve robustness against body rotation, we introduce a technique for generating artificially rotated training data using 3D keypoint reconstruction. The recognition models were trained and evaluated on a custom dataset of nine gestures, while rotation robustness was tested on a representative subset of three gestures. Results show that both models maintain high accuracy and efficiency even under moderate rotation.