Research On Face Recognition And Emotional Interaction System Of Humanoid Robot

In the process of intelligent human-computer interaction,humanoid robot not only needs to have a humanoid shape and complete the instruction of mankind,but also needs to have the imitation behavior and emotion of human.This thesis focuses on the research of face recognition and emotion interaction.The face is an important way to carry out identity authentication,moreover,it is also an important channel to transfer human emotion.On the basis of building the humanoid robot emotion interaction platform and the emotional interactive robot which is launched by laboratory,face recognition and emotional interaction is realized through the analysis and processing of facial images.Based on the existing face recognition methods,this thesis makes the deep discussion and research on feature extraction and classification,and proposes a new face recognition method and applies it to the emotional interaction system of humanoid robot.Combining with image recognition technology,voice conversation technology and physical control technology,the face recognition technology is regarded as a key point to achieve the humanoid robot emotional interaction system.The main work and innovation are as follows:(1)Due to the non-rigid face of human face and the influence of light and pose change,there is a lack of effective description method.Considering the shortcomings of existing Gabor algorithm and local binary model(LBP),this thesis presents a new feature description algorithm based on two algorithms: G-LBP.In order to further improve the stability and accuracy of the system,the face image is projected into the entropy space,and the concept of the variance cross projection entropy is defined.The entropy space and the texture space form a complementary pattern,and the face image is described from different angles.The experimental results show that the G-LBP fusion feature can effectively reduce the redundancy between the feature data while ensuring that the effective decision information is not lost.The defined variance cross projection entropy compensates for the missing information of the variance projection entropy in the face description feature.In the final classification decision,this method achieves a better recognition performance in the condition of local occlusion,expression changes,light and small posture changes,is proved to be an effective face recognition method.(2)The emotional interaction system is studied and realized based on humanoid robot.In this thesis,combined with expression,gender,voice,dialogue and other related technologies,we study the mature face recognition method and realize the humanoid robot emotional interaction system.Through real-time expression interaction,voice dialogue judgment and physical action coordination method,the communication between humanoid robot and person is realized,and the usability of emotional interaction in experimental environment is tested,which is of great significance to promote the research and development of emotional interaction.