摘要
提出一种基于动作能量图的独居老人室内异常行为识别方法。首先用背景减除法对目标进行前景提取,然后提取人体运动目标的动作能量图,对提取后的动作能量图用傅里叶描述子进行边界描述并提取主傅里叶描述子。提出一种基于最近邻分类器的分层识别算法,同时提出一种动态能量描述子DED,将主傅里叶描述子和结合动态能量描述子DED相结合,用分层识别算法对异常行为进行识别。实验表明,该方法简单实用,对独居老人室内跌倒的正确识别率高达97%。
In this paper,we propose a method for identifying the indoor abnormal behaviours of the solitary oldies which is based on action energy image.First,we use the background subtraction method to extract the foreground of target.Then we extract the action energy image of the body moving target,and describe the boundary of the extracted action energy image by using Fourier descriptor followed by extracting the main Fourier descriptor.In this paper we propose a nearest neighbour classifier-based hierarchical recognition algorithm.Meanwhile we propose a dynamic energy descriptor DED.By combining the main Fourier descriptor with the dynamic energy descriptor DED,we identify the abnormal behaviours using hierarchical recognition algorithm.Experiments show that the method is simple and practical,its correct rate of identifying the indoor falling of solitary oldies reaches 97%.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第2期188-190,共3页
Computer Applications and Software
基金
国家自然科学基金项目(1103123)
关键词
异常行为识别
动作能量图
分层识别算法
动态能量描述子
Abnormal behaviour identification
Action energy image
Hierarchical recognition algorithm
Dynamic energy descriptor
