Real-time intelligent generation of human behaviours for interactive virtual environments

从艺术和娱乐到科学模拟、教育和服务业，交互式虚拟环境在许多应用中都在迅速增长。人类角色在大多数这些应用中扮演着重要的角色。在历史上，角色动画是通过传统的关键帧技术或最近的运动捕捉技术来完成的，用于更复杂的动作。无论采用何种技术，交互式或非交互式多媒体应用中的大多数动画行为都是预先录制或预先制作的动画序列的简单回放。系统不能产生真实的行为限制了应用程序通过创建适当的内容来实时响应用户交互的能力。即使在动画电影这样的非交互式应用中，智能动画软件的支持也是最低限度的，角色动画过程的自动化将减少制作成本和工作量，并为动画师提供更有效和更系统的方式来生成内容。这项提议旨在创建一个框架，为角色动画系统增加知识和智能。这将通过行为实验和建模来实现，以了解常见的人类动作，包括但不限于面部表情、简单的全身动作和更复杂的动作，如舞蹈。研究的下一步将是开发一种软件框架，该框架使用现代系统的计算能力，并允许使用行为模型，以便以现实和感知有效的方式实时创建这些动作。最后，我们使用该框架创建样例应用程序，特别是教育虚拟环境以及计算机游戏和动画电影。拟议的研究的主要贡献将是创建动态动作的能力，免于使用昂贵的动作捕捉设备和艺术家预先录制大量动画，通过感知有效和个性化的虚拟角色来实现更逼真的环境，以及更好地理解人类行为，如情绪状态对面部和全身动作的影响

Interactive virtual environments are rapidly growing in many applications from arts and entertainment to scientific simulation, education, and the service industry. Human characters play a significant role in most of these applications. Historically, character animation is done by the traditional technique of key-framing or recently motion capture for more complicated movements. Regardless of the technique, the majority of animated behaviours in interactive or non-interactive multimedia applications are simply play-back of pre-recorded or pre-animated sequences. This inability of the systems to generate realistic behaviours limits the applications in terms of their ability to respond to user interactions in real-time by creating appropriate content. Even in non-interactive applications like animated films, there is minimum support from an intelligent animation software, and automating the process of character animation will reduce the production cost and effort, and provides animators with a more effective and systematic way to generate content. This proposal aims at creating a framework that adds knowledge and intelligence to character animation systems. This will be achieved through behavioural experiments and modeling to understand common human actions including but not limited to facial expressions, simple full-body movements, and more complicated actions such as dancing. The next step of the research will be to develop the software framework that uses the computational power of modern systems and allows using the behavioural models in order to create those actions in real time in a realistic and perceptually valid way. Finally we use the framework to create sample applications particularly educational virtual environments and also computer games and animated films. Major contributions of the proposed research will be the ability to create dynamic actions, freedom from pre-recording massive amount of animation using expensive motion capture equipment and artists, more realistic environments through perceptually valid and "personalized" virtual characters, and better understanding of human behaviours such as the effect of emotional state on facial and full-body actions