Research On Human-robot Interaction System Based On Emotion Perception And Expression Of NAO Robot

At present,human-robot intelligent interaction is a research hotspot at home and abroad.Intelligent interaction mode will redefine the relationship between human and robot,and make people's life change greatly.In human-robot interaction,if robots can communicate with human beings emotionally,and have the ability to observe,understand and generate various emotional features like human beings,it will make human-robot interaction more natural.Emotional information is obtained mainly through facial expression,semantic intonation,body posture and other aspects.In this paper,based on the application platform of robot Nao,the deep learning and robot are combined to build a human-robot interaction system based on depthwise separable convolution,which can launch real-time body emotion feedback according to facial expression recognition.The main innovative research work is as follows:(1)In terms of human-robot interaction communication,the distributed structure of client-server of human-robot interaction system is adopted,and the communication mode of server-socket-LAN-socket-client is realized,which completes the receiving and processing of interactive information and increases the real-time performance of the system.(2)In the construction and test of convolution neural network model,the feature extraction and classification of facial expressions(fear,surprise,happy,neutral,sad and angry)are carried out by using the depthwise separable convolution algorithm.The results show that the prediction accuracy of fer2013 facial expression test set can reach 71.1% and the average prediction time of a single expression picture is within 20 ms by using the trained network model.(3)In the design and verification of robot emotional expression,according to the range of joint freedom and the color distribution of LEDs in the eyes of Nao robot,six kinds of emotion actions corresponding to facial emotions are designed for Nao robot.Through the test,the recognition rate of body emotions reached 60%.(4)In the test and evaluation of human-robot interaction system,the interaction mode of human-robot interaction was shown,and the real-time interaction effect of robot was tested.In each group of interaction experiments,the prediction results of 10 consecutive frames were statistically analyzed.The experiment shows that the average total interaction time is 0.079 ± 0.010 s,the expression score of the continuous tests is relatively stable,and the robot's emotion prediction reliability to human is high.In this paper,the human-robot interaction system of robot emotion perception and expression is studied,which can be used in the fields of rehabilitation medicine and emotional care based on human-robot intelligent interaction.It is an important development trend of human-robot interaction intelligence in the future.