Development of AI-Enabled Sign Language Predicting Glove Using 3-D Printed Triboelectric Sensors

Abstract

Triboelectric energy harvesters and sensors are emerging as a desirable approach for energy harvesting and sensing. This study investigates several 3-D printed materials for the development of triboelectric nanogenerators (TENGs) and sensors. Our initial research focus is on establishing the most effective flexible 3-D printed materials for contact separation mode (CSM) TENGs. This includes a thorough examination of materials such as polyamide 6,6 (PA6,6), Vero clear, copper-coated polylactic acid (Cu-PLA), polycarbonate (PC), and acrylonitrile styrene acrylate (ASA) as well as several commercially available triboelectric negative materials. The best combination of 3-D printing (3DP) PA6,6 and Veroclear yields a maximum open circuit voltage ( $V_{\mathrm {OC}}$ ) of 63 V and an instantaneous current of $0.8~\mu $ A. With extension based on this foundation, we investigated the use of polyamide PA6,6 as a triboelectric sensor for sign language interpretation. A novel approach is adopted by integrating 3DP PA6,6 strip with an aluminum electrode onto a glove which captures the subtle movements of fingers involved in sign language. Output generated by this TENG is processed through an Arduino microcontroller which is provided with the data for alphabets A-J with high consistency and repeatability. After detailed preprocessing of the data generated by sensors, convolutional neural network (CNN), long short-term memory (LSTM), and gated recurrent unit (GRU) machine learning algorithms are utilized to train the data. Comparison analysis of these algorithms displays the highest training and validation accuracy of 99% from LSTM. These results showcase the potential of flexible 3DP Triboelectric sensors in enhancing communication for the hearing impaired.