多任务环境中基于情感的拟人Agent面部运动的实现  

Implementation of emotional facial actions for human-like agents in a multi-tasking environment

作　　者：杨斌[1] 周芳芳[1] 贾培发[1] 

机构地区：[1]清华大学计算机科学与技术系,智能技术与系统国家重点实验室,清华信息科学与技术国家实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2009年第8期1224-1228,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(60433030)

摘　　要：为解决在多任务环境中,拟人Agent多种面部运动自主综合表达的问题,该文首先结合Sloman的分层体系架构和Damasio的机体标志思想,设计了一种支持多任务环境的拟人Agent面部运动的体系结构。接着分析了拟人Agent的面部运动种类及成因,定义了面部运动的优先级,提出了冲突的协调方法,并将此方法应用到体系结构的决策控制中。最后根据提出的拟人Agent感情状态的形式化描述方法,实现了基于情感的面部运动序列的自主生成系统,解决了拟人Agent多种面部运动同时表达的矛盾,并实现了个性化的情感表达输出。Automatic synthesis of combined facial actions, such as facial expressions and movement of lips, eyes and eyelids for human-like agents, is still quite difficult. The current approach calculates combined facial actions of human like agents using a uniform model based on the anatomical structure of the human face. An algorithm is presented for multi-tasking environments that combines Sloman ＇s cogAff and Damasio＇s Somatic marker mechanisms. Then, priorities are defined for various facial actions with a method to resolve conflicts between different facial actions. An emotion model for human-like agents is then given which automatically generates combined facial actions and overcomes the conflicts between the various facial actions for personalized emotion-based expressions.

关 键 词：拟人Agent 人工情感 多任务环境 个性化 综合面部运动 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]