Research On Emotional Robot Based On Vision And PAD Model

In recent years,emotional robots have developed rapidly,occupying a very important position in the field of robots.Its emotional recognition and emotional expression capabilities make humans and robots no longer communicate mechanically,but are as harmonious as humans.Naturally,this requires robots to be increasingly capable of recognizing human emotions and understanding human emotions and conveying various emotions.This article aims to build a human-machine emotional interaction system by establishing an emotion model,combining image recognition technology and expression action control technology,and using robots to make different expressions as emotion expression methods,so that the robot can recognize human faces and reproduce human facial expressions.And the function of expressing the emotion inside the robot.During the design process,the following studies were mainly carried out:(1)Extract the Haar features of the face from the preprocessed face image and establish the Ada Boost cascade classifier,use the cascade trainer in Open CV to train and test on the face database used in this article,and combine several kinds of faces The detection algorithms are compared experimentally.The face detection algorithm used in this paper has the highest comprehensive performance in terms of detection rate and detection time,and can meet the needs of real-time detection.(2)Aiming at the problem of PCA's poor robustness in extracting features of light and facial expression changes,an LBP local binarization operator is introduced in the PCA-based face recognition algorithm,first extracting the hierarchical LBP part of the face image Feature and global features,then feature stitching and PCA dimensionality reduction,training these features extracted from the face sample set,and then extracting the features of the test face image in the same way and inputting the trained model to obtain the recognition result.Compared with the traditional algorithm,the improved algorithm not only has a higher face recognition rate than the PCA-based face recognition algorithm,but also the recognition speed can meet thesystem requirements.(3)The research compared the effects of traditional machine learning algorithms(PCA + LBP + SVM),VGG-16 network,Res Net network and Mobile Net-V2 network in facial expression recognition,and found that the recognition effect of convolutional neural network is better than traditional machine learning.The algorithm,and the recognition speed of the lightweight network is much faster than the other two networks when the recognition accuracy is slightly reduced,which can meet the real-time expression recognition required by the system.(4)Based on the PAD emotion model,the emotion model of the emotion robot is built.The labeling method of external stimulus vectors and the use of GRNN network training to obtain emotional autonomous responses are introduced.The emotion sensing vector is modified by personality,mood,and deep memory.The vectors are discretely classified to obtain specific robot emotions.Finally,human-machine emotion interaction experiments are designed to simulate robot emotion changes.Experiments prove that under this model,the robot can simply simulate the changes of human emotions.(5)The principle of emotion robot facial expression is analyzed,and the ESA system is established on the basis of the FACS system,meanwhile the emotion robot interactive system is built based on the MATLAB experimental platform,and the visual part and PAD emotion model are integrated into the experimental machine based on the Frizs expression robot.The interactive experiment not only realizes the functions of the robot to recognize human faces and reproduce facial expressions,but also enables simple human-machine emotional interaction.The research in this paper will enable the robot to analyze the user's emotions in real time during the communication with the user,and can express emotions autonomously during the interaction process,which lays some foundation for the subsequent technical research of emotional robots.