Plastic-Optical-Fiber-Enabled Smart Glove for Machine-Learning-Based Gesture Recognition

Gesture recognition has always been an important research direction in the field of human–computer interaction. In this article, a wearable gesture recognition system based on a D-shaped plastic optical fiber (POF) curvature sensor is proposed and experimentally studied. A highly bend-sensitive D-shaped POF curvature sensor is made and integrated into a five-channel signal acquisition system on a printed circuit board (8 × 4.5 cm
²
), which is embedded into an elastic glove to collect fingers’ movement data. A total of 13 gestures and 11 grasping actions are defined, and the gesture data, the grasping action data, and the gesture data mixed with grasping action data are normalized, calibrated, and imported into a support vector machine (SVM) classifier based on the Gaussian kernel function and feedforward neural networks (FNN), respectively. The recognition accuracy based on the SVM for the 13 gestures and 11 grasping actions reaches 99.8% and 97.7%, respectively. The recognition accuracy of 13 kinds of gesture data mixed with 11 kinds of grasping action data based on the Gaussian kernel function in the SVM classification model and FNN is 98.9% and 99.4%, respectively.