Abstract: In order to simulate the dynamic transformation process of emotion effectively and detect the hidden sentiment of individuals in the interaction process, so as to improve the intelligence of virtual interaction process, a Hawkes process was used to simulate the process of individual emotion generation and decline firstly. Then the individual's expression was detected by the hidden Markov model and mapped to the dimensional space of the emotional model. Finally, the sentiment of the individual was derived by comparing the emotional states that observed after and before the event stimulus. In the case when an individual hidden the emotional states, a strategy, continually appending positive stimulation to the individual, was introduced to activate the emotional expressions, and adopted the negative as the default sentiment for the unsuccessful detection of the emotional expressions. Experiments show that, this method can effectively calculate individual's hidden sentiment and improve the intelligence of interaction.