基于GRU神经网络的智能家居用户行为识别方法  

A Smart Home Users Behavior Recognition Method Based on GRU Neural Network

作　　者：高建清[1] 陈星娥[1] GAO Jianqing;CHEN Xing'e(College of Electronic Information Science,Fujian Jiangxia University,Fuzhou Fujian 350108,China)

机构地区：[1]福建江夏学院电子信息科学学院,福建福州350108

出　　处：《莆田学院学报》2020年第5期75-79,共5页Journal of putian University

基　　金：福建省中青年教师教育科研项目(JAT170620);福建省本科高校重大教育教学改革研究项目(FBJG20180310);福建省科技厅引导性项目(2019H0040)。

摘　　要：为了利用传感器数据对智能家居用户行为进行识别,提出了一种基于GRU神经网络的识别方法。首先,将数据集中的行为和传感器数据进行预处理,实现数值化和离散化;其次,按照行为的触发时间点对传感器数据进行合并处理;最后,设计出一种行为识别神经网络模型对输入的33维特征的传感器数据进行识别,输出相应的行为编号。实验结果表明,该神经网络模型的识别效果比一些传统方法和RNN神经网络模型的识别效果好。In order to identify smart home users behavior by using sensor data,this paper proposes a recognition method based on GRU neural network.Firstly,the activities in the datasets and sensor data are preprocessed to realize numerical and discrete processing;secondly,sensor data are combined according to the trigger time point of the activities;finally,an activity recognition neural network model is designed to recognize the 33-dimensional characteristics of the input sensor data and output the corresponding behavior numbers.Experimental results show that the neural network model recognition method is more efficient than some traditional methods and RNN neural network model.

关 键 词：智能家居 行为识别 GRU 神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]