项目名称: 融合尺度行为模式与环境上下文的家庭服务模式获取研究
项目编号: No.61203341
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 尹建芹
作者单位: 济南大学
项目金额: 24万元
中文摘要: 本项目在智能空间平台下研究基于人体行为表示的家庭服务模式获取方法,为家庭服务机器人或其它执行设备提供服务依据。一方面,提出基于多尺度行为模式和环境上下文的人体行为模式表示方法:提取人体轨迹,利用高斯混合模型将其转换为关键点序列,构建事件直方图,以事件直方图为基础进行大尺度行为模式表示;提取人体的轮廓特征表征人体姿态,提取时空特征点特征表示细节运动,综合姿态和细节运动表示小尺度行为模式;以人体动作结合物品特征进行物品表示,得到环境上下文。另一方面,构建基于行为模式的服务模式获取模型:分别对大尺度行为模式、小尺度行为模式和环境上下文利用主题模型进行建模,将视频信息转换为主题比例,并基于该比例特征设计分类器完成服务模式的选择。本项目旨在达到利用人体行为模式获取服务依据的目的,提高机器人主动式智能化的服务水平,对解决现阶段及未来长时间段内大量空巢老人的照料问题具有重要意义。
中文关键词: 服务机器人;智能空间;尺度行为模式;环境上下文;
英文摘要: The project researches the acquisition of family service mode based on human activity representation in intelligent space, so that the acquired results can provide the basis for the service robots or other actuators. On one hand, multi-scale activity patterns and environmental context are put forward to represent the human activity as follows: The trajectories are extracted and the Gaussian mixture models are used to cluster the trajectories into key point sequence. Based on the sequence, the event histogram is constructed to represent the large-scale pattern. The human contour features and the space-time points are used to represent small-scale pattern. The object is represented based on the space-time points and the features of the object, and then the environmental context is obtained. On the other hand, as shown below, we construct the service acquisition model based on the activity pattern: multi-scale pattern and environmental context are modeled by topic models separately, and the video information can be converted to topic proportions, based on which classifiers will be designed to acquire service pattern. The project aims to provide service basis for the robot and other actuators, so as to improve the active ability and intelligent levels of the service. It is also important for the caring of the empty
英文关键词: Service robot;intelligent space;scale behavior pattern;environmental context;