基于不确定性推理的活动识别方法研究  

Activity Recognition Method Based on Uncertainty Reasoning

作　　者：戴丹 管有庆[1] 龚锐 DAI Dan;GUAN You-qing;GONG Rui(School of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学物联网学院,江苏南京210003

出　　处：《计算机技术与发展》2022年第1期7-12,共6页Computer Technology and Development

基　　金：江苏省高校自然科学研究计划项目(05KJD520146)。

摘　　要：活动识别已成为智能家居领域的研究热点,目前国内外有关活动识别方法的研究有很多,研究人员提出了不同的方法来进行活动建模和识别,可分为数据驱动方法和知识驱动方法。数据驱动方法容易受到维数的限制,并且需要大量的数据集来训练出活动模型。目前在有关活动识别研究的方法中缺少一种既能够考虑到异构数据之间的知识共享,又能够考虑到活动的不确定性的方法。该文将D-S理论(Dempster-Shafer theory,证据理论)和本体推理结合起来,在改进的证据合成规则的基础上提出了ER-OT(evidential reasoning-ontology,证据-本体推理)算法,解决了活动中的不确定性和推理结果之间的冲突。算法首先在加权分配的思想上按重新定义的冲突系数对证据合成规则进行改进,在推理时推理机将推理信息同时输入到Jena本体推理和改进的证据推理模块,然后将推理结果按改进的证据合成规则进行合成得到最终的推理结果。实验结果表明,与现有的马尔可夫逻辑网络算法和传统的本体推理算法相比,该算法提高了不确定性活动的识别准确率。Activity recognition has gradually become a research hotspot in the field of smart home.There are currently many researches on activity recognition methods at home and abroad.Researchers have proposed different methods for activity modeling and recognition,which can be divided into data-driven methods and knowledge-driven methods.Data-driven approaches are prone to dimensionality limitations and require large data sets to train active models.There is a lack of method that takes into consideration not only knowledge sharing between heterogeneous data but also uncertainty of activities in the research of activity recognition.In this paper,evidence theory and ontology reasoning are combined.Based on the improved rules of evidence synthesis,an algorithm of evidence ontology reasoning(ER-OT)is proposed to solve the conflict between uncertainty and reasoning results in activities.Firstly,in the idea of weighted distribution,the rules of evidence synthesis are improved according to the redefined conflict coefficient.In reasoning,the inference engine inputs the reasoning information into Jena ontology reasoning and improved evidence reasoning module at the same time,and then the reasoning results are synthesized according to the improved rules of evidence synthesis to get the final reasoning results.Experimental results show that compared with the existing Markov logic network algorithm and the traditional ontology reasoning algorithm,the proposed algorithm improves the recognition accuracy of uncertain activities.

关 键 词：活动识别 本体推理 D-S理论 智能家居 不确定性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]