WEARABLE SENSOR TECHNOLOGIES FOR AUTOMATIC SIGN LANGUAGE RECOGNITION: MODERN APPROACHES AND THE IMOVOX SMART GLOVE

Abstract

This thesis analyzes wearable sensor technologies used for automatic sign language recognition, their advantages and practical limitations, and examines the experimental prototype of the IMOVOX smart glove. In modern systems, the integration of flex sensors, inertial measurement units (IMUs), microcontrollers, and machine learning algorithms enables the simultaneous recording of hand shapes and spatial movements. The IMOVOX prototype was developed using five flex sensors, an MPU6050, and the ESP32 platform. A total of 900 trials were conducted with 10 participants across 30 basic signs, achieving an overall accuracy of 83.1% and an average response time of 452 ms. The results showed that static signs were recognized more consistently than dynamic signs. The study indicates that sensor fusion, individual calibration, and TinyML-based classification offer promising opportunities for the further development of local assistive devices such as IMOVOX.