论文部分内容阅读
对人行为的感知和分析是家庭监护服务机器人系统中的关键环节.本文在概率框架下提出一种基于抽象隐马尔可夫模型的人运动行为识别方法.室内环境中人的运动具有层次化特性且各层次具有抽象马尔可夫决策过程的性质,因此采用具有级联形式的抽象隐马尔可夫模型建模人的运动.使用期望最大化算法分别学习抽象隐马尔可夫模型的观测模型和状态转移模型,采用具有较高计算效率的Rao-blackwellised粒子滤波近似推理方法识别人运动的时空序列.实验数据采用视觉跟踪与定位系统获得的人体运动轨迹,运用该法训练并识别多种室内运动模式,结果证明本文方法的有效性.
Perception and analysis of human behavior is the key link in the family of guardianship robot system.This paper presents a method of human motion identification based on abstract hidden Markov model in the framework of probability.Human movement in the indoor environment has the characteristics of hierarchical And each layer possesses the properties of abstract Markov decision making process, the abstract motion of hidden Markov model with cascade is used to model the human motion.We use the expectation maximization algorithm to learn the observation model and state of abstract hidden Markov model Transfer model, the Rao-blackwellised particle filter approximate reasoning method with higher computational efficiency is used to identify the spatial-temporal sequence of human motion.The experimental data adopts the trajectory of the human body obtained by the visual tracking and positioning system, and uses this method to train and identify a variety of indoor exercise modes , The result proves the validity of the method in this paper.