Research On Emotion Interaction In Human-computer Interaction Based On Emotional Cognitive Evaluation Theory

Emotions play an important role in the daily lives of people study and work, and emotion is the biggest difference human beings and current intelligent machines. Continuous improvement of cognitive psychology, cognitive neuroscience and emotional cognitive evaluation theory help the emotional robot research in the field of artificial intelligence. Computers and artificial intelligence research scholar increasingly find intelligent machines only with the logic of computing power is not enough, should also be allowed to emotional competence (emotional identification and emotional expression).This paper first describes the status and significance of intelligent robots and artificial emotion, then describes the basic theoretical knowledge and technology related to this topic, and then describes the emotion model based on emotional cognitive appraisal theory, last describes affective cognitive decision-making(ACD) of the affective-cognitive(AC) model and by improving the Q-learning algorithm to achieve ACD, the traditional Q-learning algorithm does not consider itself to achieve certain goalsself-incentive mechanism to consider only the reward from the external environment, and improved Q-learning is used to both reward and punishment mechanism, since the reward incentive mechanism and the external environment, while using the BP neural network in the Q-learning Q function.ACD is divided network, that each decision-making by its sub-decision and primitive skill, and each decision is able to start or stop its sub-decisions. Network layered on ACD in order to resolve the decision-making dimension, its essence is the emotional and cognitive decision-making be broken down, ACD on different levels have their own goals, each low-level ACD only in a smaller state space to solve, learn the strategy of low-level emotional cognitive decision-making can be a high-level ACD is called directly, thus speeding up the decision-making to solve the speed.Designed based on emotional cognitive evaluation theory and emotional decision-making ability, emotional intelligence body, it is composed of the observation module, module emotional, cognitive modules, emotional, cognitive decision-making module and emotional behavior module and other modules. Emotional agent get information from outside by attention model, then the appropriate decision-making through the emotional cognitive model in the emotional cognitive decision-making model, last emotional agent perform the appropriate emotional behavior, emotional behavior decision-making results in feedback to the emotional agent.Finally, the use of open source OpenCV and Access on VS2010program a emotional agent that able to identify a simple facial expressions by using Ekman’s facial expression coding system. Emotional agent have initially in human-computer interaction with emotional decision-making capacity, while improved Q-learning algorithm was validated on matlab. The simulation experimental results show that emotional agent’s emotional behavior of the reaction is in line with the laws of human emotional behavior, based on emotional cognitive evaluation theory of ACD method is effective and achieved the desired results.