基于毫米波雷达与情感神经网络的室内人员跌倒检测算法

Indoor Fall Detection Algorithm Based on Millimeter Wave Radar and Emotional Neural Network

为了实现室内人员的跌倒检测,提出一种基于毫米波雷达的跌倒检测(Human Fall Detection,HFD)算法。首先,利用环境杂波滤除算法去除雷达数据的杂波信号;其次,在三维笛卡尔坐标系中利用参考质心聚类算法和躯干特征算法得到人物质心状态向量;最后,利用连续径向基情感神经网络(Continuous Radial Basis Emotional Neural Network,CRBENN)推断算法判断人物是否跌倒。实验结果表明,该算法能够在多种动作中实现人物跌倒判断及其定位,其检测准确率为99.28%,数据处理时间不超过35.8 ms,体现出了较好的准确性和实时性。

In order to realize indoor fall detection,a human fall detection(HFD)algorithm based on millimeter wave radar is proposed.Firstly,environmental clutter filtering algorithm is used to filter the clutter signal of radar data.Secondly,reference centroid clustering algorithm and trunk feature algorithm are employed to obtain the state vector of centroid in three dimensional Cartesian Coordinate System.Finally,inference algorithm of continuous radial basis emotional neural network is applied to judge whether the person falls.The experimental results show that the algorithm can judge and locate people's fall in many different actions with the detection accuracy of 99.28%and processing time of less than 35.8 ms,demonstrating good accuracy as well as real-time performance.