Research On Key Technology Of Affective Computing

Human being not only has the ability of logical reasoning, but also other abilities such as producing, controlling and expressing emotions. It has been a great challenge for the researchers in the field of AI to endow computers with emotions since the success in enabling them of logical reasoning ability. In recent years, the technology of affective computing has been paid great attention to by academic and business communities since it was proposed by Picard at MIT in 1997. The ultimate goal of research in affective computing is to let computers have emotions, so that we can establish people-oriented environment in human-computer interaction (HCI) in the future.This paper will discuss the studies I have conducted on some key problems in related field of affective computing. I did research mainly in the mixed emotional classification method based on 2-D coordinate, emotional approach-withdrawal model based on emotion-recognition, and emotional interaction model based on dynamic Q-learning arithmetic. We design an application system to monitor user emotions in the English text of Email by using emotional agent's characteristics. At last, tests and evaluation of the system were carried out with the help of technology acceptance model (TAM). The following are the main contents and novelty of the research:(1) Research on the method of mixed emotional classification based on 2-D coordinateI propose a method of mixed emotional classification based on 2-D coordinate in the context of existing theories of discrete emotion and dimension emotion. This method is used to describe emotions in 2-D coordinate. This one is similar to emotional dimension theory because they are used to describe emotions in 2-D planar. However, the two have some differences. The dimension theory describes emotion as an isolated point, while the mixed emotional theory describes emotions as a set. This method is in accordance with the changing of complex emotions. At last, I tested the method of mixed emotional classification by modeling speech emotion. In the experiment, we build up the linear model of speech emotion with acoustics features of six basic emotions. The experimental results have proved the effectiveness because texts in EmailThis is to develop and extend the application of affective computing. Email is one of the most widely used applications on the Net. We design the emotion monitoring system based on Email texts, which is good for harmonious environment of HCI. The sentences in Email come from Commonsence. We chose some sentences from Commonsence. Four structure models are created according to these sentences. At last, we realized emotion monitoring of texts with programm launguage. In the procedure of system implementation, we mark the textal emotions with colours. This method is one novelty of the paper.(5) Research on TAM about emotional interface agent in Email for usersThe model is used to predict and evaluate part (4) mentioned above in this paper. TAM is technology acceptance model. We introduced the variable of user real-time emotion in intrinsic TAM. We proved that emotions play important roles in the procedure of accepting some new technology for this variable. Therefore, we studied many problems, such as questionnaire survey, reliability analysis, validity analysis, regression analysis and hypothesis testing after we put forward the variable. At last, this new technology was welcomed by a user acceptance of 64.4%. Obviously, the new technology can bring us economic benefit and provide convenience for users with Email.Finally, the conclusion gives a summary of my research project and discusses the deficiency in the project. It also probes into the prospect of my future work in the field.