A Deep Learning-Powered TinyML Model for Gesture-Based Air Handwriting Simple Arabic Letters Recognition

This paper offers a novel deep learning (DL)-powered Tiny machine learning (TinyML) model meticulously customized to recognize Arabic hand gestures (AHGs) executed in mid-air. The principal focus of this study lies in the intricate task of precisely classifying Arabic letters through these gestures. The paper provides a comprehensive exposition of the complex dataflow architecture, encompassing the processing of gyroscope and accelerometer data to derive precise 2D gesture coordinates. The pivotal role of convolutional neural networks in the DL model is elucidated, emphasizing their outstanding performance in reaching a level of accuracy of 94.2% in classifying diverse AHDs. This accuracy showcases the model’s efficacy and robustness, underscoring its potential for real-time, practical deployments in scenarios of gesture recognition. The implications of this research extend beyond the immediate domain of Arabic letter recognition, contributing to the progress of TinyML applications in real-world gesture recognition apps.