目的人类行为理解是机器智能研究中最富有挑战性的领域。其根本问题是语义获取,即从动作推理得到人的行为,需要跨越两者之间的语义鸿沟,为此提出一种人关于日常行为知识与人体动作行为、环境信息之间的建模方法,以及可扩展的开放式结构环境—行为关系模型,基于该模型提出一种新的行为理解的渐进式认知推理方法。方法首先根据知识,建立多种特征、复合特征和行为之间的关系模型。系统根据当前的输入流,处理得到当前的特征与复合特征集,推理得到当前的可能行为集。该行为集指导处理模块,更新特征集,得到新的行为集。结果应用本文渐进式连续推理方法,系统可以把人关于日常行为的知识与人体运动、环境变化等传感器数据处理获取到的信息动态绑定,实现知识辅助的行为理解。结论提出的推理方法能连续处理长时间、同时发生的行为。 Aims Human behavior understanding is one of the most challenging areas in machine intelligence research. The basic problem is the semantic acquisition, that is to obtain the human behavior from the inference of action. It is necessary to cross the semantic gap between the two. Therefore, a method of modeling the relationship between daily behavior knowledge and human behavior and environmental information is proposed. As well as a scalable open structure environment-behavior relation model, based on which a new cognitive cognition method of behavior comprehension is proposed. The method first builds a relational model of multiple features, complex features, and behaviors based on knowledge. According to the current input stream, the system processes the current feature and composite feature set, and infers the current set of possible behaviors. This behavior set guides the processing module, updates the feature set, and gets a new set of behaviors. Results By using the method of incremental continuous reasoning in this paper, the system can dynamically bind people’s information about daily behavior and sensor data processing, such as human movement and environment change, so as to realize knowledge-assisted understanding of behavior. Conclusion The reasoning method proposed can deal with long-term and simultaneous behaviors.