摘要
基于体感网对人体动作进行识别的很多研究都是针对单人动作,很少有研究讨论双人交互动作的识别。针对双人交互动作中两人肢体行为的特点,提出了一种隐马尔可夫模型和马尔可夫逻辑网相结合的方法。其中,单人原子行为通过建立隐马尔可夫模型来进行识别,在两人交互行为的语义建模中,建立一阶逻辑知识库,并通过训练马尔可夫逻辑网来最终实现两人交互行为的决策。实验结果表明,与基于特征层数据融合的一些方法相比,该方法获得了更高的识别精度,能够有效地识别出双人交互动作。
Existing work in human activity recognition based on Body Sensor Networks(BSN)mainly focuses on recog-nizing single-user activities and lacks of discussions about two-body interactive activities. A new hierarchical recognition framework which consists of Hidden Markov Model(HMM)and Markov Logic Network(MLN)is proposed according to the characteristics of two-body interactive actions. The primitive actions of a single person are recognized by using Hidden Markov Model, and the final decision of interactive actions is made by constructing first-order logic knowledge base and employing MLN. Experimental results on the interaction dataset show that the proposed method can achieve a higher accuracy compared to other methods in activity recognition of two-body interactions.
出处
《计算机工程与应用》
CSCD
2014年第13期1-5,20,共6页
Computer Engineering and Applications
基金
国家高技术研究发展计划(863)(No.2012AA04150502)
国家自然科学基金(No.61174027)
国家科技支撑计划项目(No.2012BAK15B05-03
No.2013BAK03B01)
辽宁省高等学校杰出青年学者成长计划(No.LJQ2012005)
关键词
体感网
双人交互动作
隐马尔可夫模型
数据融合
一阶逻辑
马尔可夫逻辑网
Body Sensor Networks (BSN)
two-body interactive activities
Hidden Markov Model (HMM)
data fusion
first-order logic
Markov Logic Network(MLC)