An Improved Faster R-CNN Algorithm for Gesture Recognition in Human-Robot Interaction

摘要

Gesture recognition technology has been recognized as a communication method between human and robot systems. To our best knowledge, however, no study has been reported in open literature regarding the faster region-based convolutional neural network (Faster R-CNN) algorithm to improve the accuracy of gesture recognition. This paper proposes an improved Faster R-CNN algorithm that combines three key insights: (1) one can use Gaussian filter as image pre-processing method in order to remove image noise, (2) VGG16, compared with the residual network, is used as the feature extraction network of the improved Faster R-CNN algorithm to improve the gesture recognition accuracy, (3) five-fold cross-validation is used to evaluate the generalization performance of the algorithm. The experimental results show that an improved Faster R-CNN algorithm significantly improves mean average precision to 99.89%, which provides a better method for gesture recognition in human-robot interaction applications.