摘要
针对人体活动识别,现有的研究更多关注于传感器之间的融合,较少将研究方向转移到三轴加速度计等传感器内轴之间的相关性方面。为有效利用轴之间的相关性,提出一种基于数据融合的卷积神经网络方法。借助利用轴之间的相关性的单通道数据融合方法得到融合数据,将融合数据输入到卷积神经网络中提取特征。在WISDM数据集上的实验结果表明,该方法的准确率达到了98.80%,优于不使用数据融合的卷积神经网络方法。
Existing research on human activity recognition focuses more on the fusion of multiple sensors,with little attention to the correlation between the inner axes of the sensors such as triaxial accelerometers.To effectively use the correlation between the axes,a data fusion based convolutional neural network approach was proposed.The single-channel data fusion method using the correlation between the axes was adopted to obtain the fused data,and the fused data were inputted to the convolutional neural network to extract features.Experimental results on the WISDM dataset show that the accuracy of the proposed approach is 98.80%,which is superior to the convolutional neural network method without data fusion method.
作者
韩欣欣
叶剑
周海英
HAN Xin-xin;YE Jian;ZHOU Hai-ying(School of Data Science and Technology,North University of China,Taiyuan 030051,China;Research Center for Ubiquitous Computing Systems,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China;Beijing Key Laboratory of Mobile Computing and Pervasive Device,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China)
出处
《计算机工程与设计》
北大核心
2020年第2期522-528,共7页
Computer Engineering and Design
基金
国家重点研发计划基金项目（2016YFB1001100）
关键词
人体活动识别
三轴加速度计
数据融合
卷积神经网络
准确率
human activity recognition
triaxial accelerometer
data fusion
convolutional neural network
accuracy
