Dynamic Gesture Recognition Algorithm based on ROI and CNN for Social Robots
Jing Yang, Weihua Sheng, Guanci Yang · 2018
To eliminate barriers to communication between social robots and disabled people in listening and/or speaking, this work proposes a gesture recognition method that is based on the region of interest (ROI) and a convolutional neural network (CNN) for social robots. This method can track and recognize gestures in real time in a complex background. ROI and OpenCV are first used to obtain the dynamic gesture, which is then treated as the input of a CNN model to output a gesture feature model. Furthermore, a gesture-controlled social robot is implemented by the obtained feature model. Finally, the performance of the system is verified, and experimental results show that the proposed technology can track and recognize user gestures in real time.