Research On Human-machine Interaction Of Humanoid Head Robot Based On Artificial Emotion

Due to the vigorous development of robotics and the development of related extended functions,people putting forward higher requirements for human-machine interaction experience.Many experiments and studies have demonstrated the effectiveness of emotional expression in human-machine interactions.At present,many emotional models have emerged in the field of emotional computing.However,most existing artificial emotion models are designed for specific evaluation rules and lack a unified emotional evaluation framework.And there is less research to apply artificial emotion model to actual robots for human-machine emotional interaction.Emotion model and human-machine interaction are not closely related.Aiming at the above problems,this paper proposes a more general artificial emotion model,and integrates the model with the H&F robot-III humanoid head robot.It has the ability to express emotions and realize human-computer emotional interaction.This paper first constructs the overall framework of the artificial emotion model based on the three-dimensional emotion space of PAD(Pleasuree?Arousal?Dominance)in four parts.Emotional annotations for several common interaction events,turning these events into an array that the computer can handle.Design emotional sample collection questionnaire and collect emotional samples.A generalized regression neulral network(GRNN)is trained based on the these samples so that it can generate an autonomous emotional response to the external stimulus and generate a PAD emotion response vector.Afterwards,the personality and mood of the person are modeled,and the emotional response is corrected to make the emotional response more accurate and full of personality.The modified PAD emotion response vectors are classified to obtain the corresponding specific emotion classes.In this section,first split and expand the original classification sample set.After that,the classification features are selected to eliminate redundancies and extraneous features.Finally,training random forest classifiers to achieve emotion classification.After testing,the correct classification rate of the random forest classifier can reach more than 85%.After getting the specific type of emotion,calculate the emotional intensity.This article defines the calculation method for the internal state of the robot.And simulates the model on the MATLAB platform.Through the analysis and evaluation of the simulation results,it can be considered that the simulation results verify the validity and rationality of the model.Therefor this article defines robot expression synthesis and voice response rules.The artificial emotion model and a speech recognition module are integrated on the existing H&F robot-III humanoid head robot platform in the laboratory so that it can detect human speech and perform autonomous emotional response and emotional expression on the speech stimulus.Relying on this experimental platform,a simple human-computer interaction experiment was conducted.The research in this paper will make H&F robot-III humanoid head robots have human-like emotional responses and emotional expressions,providing a theoretical basis for further study of human-computer emotional interaction.