基于LPP和HMM的人体动作识别
首发时间:2012-06-21
摘要:人体动作序列的变化性特别强,难以找到普适的特征进行描述。本文以人体动作剪影图像序列作为输入,运用主成分分析(PCA)对数据进行预处理,并采用局部保持投影(LPP)方法从高维动作序列数据中获取动作的低维流形嵌入;然后为了充分利用动作姿态之间的时序关系,采用几种隐马尔科夫模型(HMM)实现动作识别;最后讨论了几种模型对动作识别结果的影响。实验结果表明,本文采取的方法取得了较好的识别结果,验证了本方法对动作识别的有效性。
关键词: 信号与信息处理 人体动作识别 局部保持投影 隐马尔科夫模型
For information in English, please click here
Human Action Recognition Based on LPP and HMM
Abstract:The changes between different human action sequences are obvious, and it is hard to find a universal characteristics to describe. In this paper, the human body silhouette image sequences are chosen as the input vectors, and principal component analysis (PCA) is adopted for data pretreatment. And the Locality Preserving Projections (LPP) is used to obtain the low dimensional manifold embedding of the high dimensional actions sequences.Then,in order to make the best of the information between every close pose,Hidden Markov Models (HMM) are introduced to recognize the actions.Finally discuss the effects of those models to action recognition. The experiments show that this method get a high accuracy, which fully prove that this method could describable actions effectively.
Keywords: Signal and information processing human action recognition LPP HMM
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于LPP和HMM的人体动作识别
评论
全部评论0/1000