Dynamic Gestures Recognition through a Low-cost Data Glove

This paper presents some results towards the realization of a commercial wearable device for sign language translation. In particular, we propose an entire workflow for recognizing dynamic gestures. Different techniques are analysed and compared, both to realize a database of gestures' features and to classify a performed gesture. We test the proposed method on dataset of 10 dynamic gestures, acquiring data of two IMUs embedded in a data-glove. Different classifiers achieve a 100% of accuracy, using spline fitting as feature extraction technique.