摘要
针对室内火灾情境下人员安全的有效监控问题,提出了一种基于手机多源传感器的室内火灾行人细粒度行为识别与匹配方法.借助手机内置多源传感器完成对行人当前表征行为特征的数据采集,在异常子序列探测后提取行为特征向量,利用基于关键点序列的动态时间规整(Key-DTW)算法或者相应训练成型的分类模型分别对特征各异的行为进行匹配、理解;并对不同传感器组合方式和不同设备位置的识别能力进行比较;最后,综合识别结果进而分析行人当前生理、心理、位置状态,为室内应急救援工作提供决策信息.经模拟试验验证,该方法不仅能够对行人应激性细粒度行为有较高的识别准确率,对于持久性的动作也有着很高的匹配精确性和效率.
When a compartment fire, it is impossible to monitor the safety of pedestrian effectively, a method of pedestrian’s fine-grained behavior recognition based on built-in sensors of smartphone was proposed. In this method, the multi-sensors of mobile phone were used to collect the data of the pedestrian’s characterization. After detecting the abnormal sub-sequence, the feature vectors were extracted. Then, the algorithm of Key-DTW and the models of classifying were respectively used to recognize and understand the pedestrian’s activities. Next, comparing the ability of classifying in different position of device and in various combination of smartphone sensor. Finally, analyzing the pedestrian’s current status of physiological, psychological and positional. The method will provide much valuable information for the rescue operation. The results of experiments showed that the method has higher accuracies and efficiencies of activity recognition.
作者
陈国良
曹晓祥
CHEN Guoliang;CAO Xiaoxiang(NASG Key Laboratory of Land Environment and Disaster Monitoring, Xuzhou 221116,China;School of Environmental Science and Spatial Informatics, China University of Mining andTechnology, Xuzhou 221116, China)
出处
《同济大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2019年第3期414-420,共7页
Journal of Tongji University:Natural Science
基金
国家重点研发计划(2016YFB0502105)
国家自然科学基金(41371423)
江苏省自然科学基金(bk20161181)
江苏高校品牌专业建设工程(PPZY2015B144)
