基于智能视觉的人体跌倒检测仿真  

Simulation of Human Fall Detection Based on Intelligent Vision

作　　者：肖巍[1] XIAO Wei(Changchun Normal University,Jilin Changchun 130032,China)

机构地区：[1]长春师范大学,吉林长春130032

出　　处：《计算机仿真》2020年第4期410-413,450,共5页Computer Simulation

基　　金：吉林省教育厅“十三五”科学技术项目(JJKH20170655KJ);长春师范大学自然科学基金项目(2016-010)。

摘　　要：针对传统人体跌倒检测方法准确度低,不能在人体疑似跌倒的第一时间及时检测的问题,提出基于智能视觉的人体跌倒检测方法。根据智能视觉分析技术解析人体跌倒行为,采用加速度传感器采集人体跌倒惯性特征数据并利用加速度传感器建立三轴加速坐标,对跌倒行为作出判断。在巨大的特征量集合中,运用K-L变换方法提取出准确的加速度峰值和倾角变化值,据此设置跌倒行为检测的约束条件,完成对跌倒行为的分类。采用PSO分类器优化人体跌倒检测的SVM参数,完成人体跌倒高精准度检测。实验结果表明,所提方法的检测准确度高于对比的3种文献方法,检测时间最短,能够及时检测目标个体跌倒情况,可广泛应用于现实生活中。In this article, a human fall detection method based on intelligent vision was put forward. According to the intelligent vision analysis technology, the human fall behavior was analyzed. The acceleration sensor was used to collect the inertia characteristic data of human fall, and the three-axis acceleration coordinate was established to judge the human fall behavior. In the huge set of features, K-L transformation method was used to extract the accurate peak value of acceleration and the change value of inclination angle. On this basis, the constraint conditions for human fall behavior detection was established to complete the classification of human fall behavior. Finally, PSO classifier was used to optimize SVM parameters of human fall detection, and thus to complete high-precision human fall detection. Simulation results show that the detection accuracy of the proposed method is higher than that of other methods in literatures, and its detection time is the shortest, so it can detect the situation of human fall of target in time. This method can be widely applied in real life.

关 键 词：智能视觉 人体跌倒 阈值检测 模式识别 加速度传感器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]