随着人口老龄化和科技的发展，助老机器人在家庭中的应用越来越广泛。除日常功能外，及时发现老年人所处异常状态进而对老年人实现即时救助，也成为助老机器人必不可少的一项重要智能。针对助老机器人服务对象异常行为发现这一重要问题，从位置、动作特征信息融合的角度对其进行研究。首先，对老年人的位置和动作特征进行提取，建立基于DAEI的动作特征模型；然后，利用SLAM地图的位置信息和动作图像局部二阶相对矩生成特征向量；最后，通过改进的FCM算法对特征向量进行聚类。实验结果表明了该方法的可行性。

With the population aging and the development of science and technology, the service robots for the elderly are used widely. Except for daily functions, the service robots in the family environments should have more intelligence.One of the most important abilities is to detect the unusual behaviors of clients, and call for help in time. In order to achieve this, we propose an approach for the service robot to detect the unusual behaviors the elderly in a realtime way and construct a solution to the problem of the research model. First, we get the DAEI feature model of the action feature,and then propose the locationaction features of the position and action from the location in SLAM and DAEI’s local relative moment. Finally, through the improved FCM algorithm clustering feature vectors, the experimental results demonstrate that the model is feasible and can afford a judging gist to detect the persons' unusual behaviors.