摘要
为提高基于智能手机三轴加速度传感器的人体行为分类模型准确率,提出了一种卷积神经网络人体行为识别方法,通过固定时间窗口对连续样本数据进行切割分片,构造多层神经元网络结构,优化调整关键核心参数,使用新方法训练出的人体行为识别模型平均交叉准确率可达91.7%,较其他传统机器学习算法有较大提升.
In order to improve the accuracy of human behavior recognition based on the smartphone's three-axis acceleration sensor,this paper proposes a human behavior recognition method based on convolutional neural network.This method cuts and slices continuous sample data through a fixed time window,constructs a multi-layer neuron network structure,optimizes and adjusts key core parameters.The average crossover accuracy of the trained human behavior recognition model can reach 91.7%,which is higher than other traditional machine learning.
作者
林峰
LIN Feng(Fuzhou Polytechnic,Fuzhou 350108,China)
出处
《通化师范学院学报》
2023年第12期61-66,共6页
Journal of Tonghua Normal University
基金
福建省中青年教师教育科研项目(JAT210813)。
关键词
人体行为识别
卷积神经网络
三轴加速度传感器
human behavior recognition
convolutional neural network
three-axis acceleration sensor