应用多传感器和SVM的跌倒检测系统

Fall-detecting System Based on Multi-sensor and Support Vector Machine

常见的跌倒检测方法包括基于视频、基于穿戴式和基于环境传感器三种,每种方法都有其优缺点。本文给出一种基于多传感器和支持向量机(SVM)的人体跌倒检测系统,该系统首先基于摄像头和加速度计同时获取人体运动信息,提取数据特征,然后融合特征构建特征向量,利用SVM对特征向量进行分类,最终确定跌倒与否。实验测试表明,利用多传感器进行跌倒检测可以使准确度达到99.5%,与单一传感器比较,检测性能有显著提高,误判和漏判情况明显减少。

Common fall detection methods include video-based,acceleration-based and environment sensor-based.Each method has its advantages and disadvantages.This paper presents a human fall detection system based on multi-sensor and support vector machine(SVM).The system firstly obtains human motion information based on camera and accelerometer at the same time,extracts data features,then fuses features to construct feature vectors,classifies feature vectors using SVM,and finally determines whether the human fall or not.The experimental results show that the accuracy of fall detection with multi-sensor can reach 99.5%.Compared with single sensor,the detection performance has been significantly improved,while the false positives and false negatives decrease significantly.