室内环境下老年人异常行为检测系统设计与研究

彭小英

摘 要

目前，我国老龄化进程不断加快，空巢和独居老人比例明显上升，随着年龄的增长，老年人身体机能不断退化，非常容易发生意外伤害。不少研究表明，跌倒、碰撞等异常行为是造成老人伤害的主要原因。由于老人一天中约有90%的时间都是在室内环境中度过，因此，本文主要解决室内环境下的老年人异常行为检测问题。通过分析老年人的行为活动，发现跌倒是老年人伤害事件中发生比例最高、伤害性最大的异常行为，而步态异常也会诱发跌倒等严重风险。所以，本文将跌倒这种异常行为作为研究重点，步态参数作为发现异常行为的补充，设计了一个老年人异常行为检测系统。本文的主要研究工作包括以下几个部分：

1、数据采集：本文选择阵列式柔性压力传感器作为行为数据采集子系统的感知模块，然后设计压力信号获取方案，利用多路复用器控制模拟通道实现传感器的阵列扫描，再通过分压电路和模数转换器将由于形变产生的电阻信号转换为数字信号，并使用C语言编写单片机的采集软件，最终获取行为数据，生成压力图像。

2、跌倒检测：由于不同行为事件对应的传感器压力感应状态不同，本文以压力图像中包围受力感应点的矩形框为主要分析对象，根据矩形框的属性特征确定压力感应点个数、矩形框面积、覆盖率和边长比四个特征变量，建立基于SVM的异常行为检测模型，实现站立、坐着、行走和跌倒行为的分类识别。并基于362条实测数据对模型的性能进行验证，结果显示，该模型的准确率高达95.714%，能够准确实现跌倒行为的检测。

3、步态获取：根据传感器的实际布置区域建立对应位置的地面传感坐标系，利用传感单元点的已知位置，推导步态参数计算规则，实现步态信息的获取。

本文设计的异常行为检测系统能够减少对老人的生活干预，保证老人的生命安全，为老人的异常行为检测提供了新的研究方向，具有重要应用价值。

关键词：异常行为；步态参数；行为数据；支持向量机

Abstract

At present, my country's aging process is accelerating continuously, and the proportion of empty nesters and solitary elderly has increased significantly. With the increase of age, the physical functions of the elderly are deteriorating continuously, and accidental injuries are very likely to occur. Many studies have shown that abnormal behaviors such as falls and collisions are the main causes of injuries to the elderly. Since the elderly spend about 90% of their time in the indoor environment, this article mainly solves the problem of detecting the abnormal behavior of the elderly in the indoor environment. By analyzing the behavioral activities of the elderly, it is found that falls are the abnormal behavior with the highest proportion and the most harmfulness among the elderly, and abnormal gait can also induce serious risks such as falls. Therefore, this article takes the abnormal behavior of falling as the research focus, and gait parameters as a supplement to discover abnormal behavior, and designs a detection system for the abnormal behavior of the elderly. The main research work of this article includes the following parts:

1. Data collection: In this paper, an array-type flexible pressure sensor is selected as the sensing module of the behavioral data collection subsystem, and then a pressure signal acquisition scheme is designed. The multiplexer is used to control the analog channel to realize the sensor array scanning, and then through the voltage divider circuit and The analog-to-digital converter converts the resistance signal generated by the deformation into a digital signal, and uses C language to write the acquisition software of the single-chip microcomputer, and finally obtains the behavior data and generates the pressure image.

2. Fall detection: Due to the different pressure sensing states of the sensors corresponding to different behavior events, this paper takes the rectangular frame surrounding the force sensing point in the pressure image as the main analysis object, and determines the number of pressure sensing points and the rectangular frame according to the attribute characteristics of the rectangular frame Four characteristic variables of area, coverage and side length ratio are used to establish an abnormal behavior detection model based on SVM to realize the classification and recognition of standing, sitting, walking and falling behaviors. And based on 362 measured data to verify the performance of the model, the results show that the accuracy of the model is as high as 95.714%, can accurately detect the fall behavior.

3. Gait acquisition: The ground sensing coordinate system corresponding to the position is established according to the actual layout area of the sensor, and the known position of the sensor unit point is used to derive the gait parameter calculation rule to achieve gait information acquisition.

The abnormal behavior detection system designed in this paper can reduce the intervention of the elderly's life, ensure the safety of the elderly's life, and provide a new research direction for the elderly's abnormal behavior detection, which has important application value.

Key words：Abnormal behavior；Gait parameters；Behavioral data；Support Vector Machines