Development of an OTDR-Based Hand Glove Optical Sensor for Sign Language Prediction

This article demonstrates a wearable optical fiber-based sensing glove for predicting hand gestures for sign language. The sensors were created using macrobends on single mode fiber (SMF) for each finger to simultaneously monitor the change in bending radius due to the change in position of each finger by creating different hand gestures. The glove sensing capability is based on the change in macrobending loss due to a change in bending radius. The system includes an optical time-domain reflectometer (OTDR), optical fiber, and hand glove. The functionality and performance of the glove were thoroughly evaluated on different participants for repeatability. We retrieved hand gesture information using data from each finger macrobend loss through the sensor’s response on OTDR—the developed glove acquired real-time raw data on alphabet and numerical hand postures. A set of unique features were extracted from the acquired raw data. The selected features were used to classify hand gestures using the ensemble classifier. The performance of this classifier was optimized using the training dataset (70%), and the performance of this trained algorithm was evaluated to recognize hand gestures using a testing dataset (30%). Compared to some existing methods, the proposed method enhances sign language recognition accuracy and improves tracking of hand gesture movement. This article presents statistical results with 93.57% classification accuracy on test data representing good dynamic gesture recognition scenarios.