An EMG dataset for Arabic sign language alphabet letters and numbers

Amina Ben Haj Amor; Oussama El Ghoul; Mohamed Jemni

doi:10.1016/j.dib.2023.109770

An EMG dataset for Arabic sign language alphabet letters and numbers

Data Brief. 2023 Nov 4:51:109770. doi: 10.1016/j.dib.2023.109770. eCollection 2023 Dec.

Authors

Amina Ben Haj Amor¹, Oussama El Ghoul¹, Mohamed Jemni¹

Affiliation

¹ Research Laboratory LaTICE, University of Tunis, Tunisia.

Abstract

Nowadays, surface electromyography (sEMG) is evolving as a technology for hand gesture recognition. Detailed studies have revealed the capacity of EMG signals to access detailed information, particularly in the classification of hand gestures. Indeed, this advancement emerges as an interesting element in refining the recognition and interpretation of sign languages and exploring deeper into the phonology of signed languages. Aligned with this advancement and the need for a reliable and mobile sign language recognition system, we introduce a specialized sEMG dataset, acquired using the Myo armband. This device is adept at capturing recordings at frequencies of up to 200 Hz. The dataset focuses on the 28 letters of the Arabic alphabet and 10 digits using hand gestures, with each gesture captured into 400 frames. This considerable collection of 18,716 samples was achieved with the cooperation of three contributors, providing a varied and comprehensive range of gestural data.

Keywords: Arabic sign language; Gesture recognition; Myo Armband; Surface Electromyography sEMG.