一种基于改进深度迁移学习的睡姿识别算法  

An Improved Deep Transfer Learning Algorithm for Sleeping Posture Recognition

作　　者：王述 许海棋 陈中豪 任佳[1,2] 

机构地区：[1]浙江理工大学自动化系,杭州浙江310018 [2]中国计量大学浙江省智能制造质量大数据溯源与应用重点实验室,杭州浙江310018

出　　处：《工业控制计算机》2023年第5期78-80,共3页Industrial Control Computer

基　　金：浙江省公益技术研究项目(LGG20F030007);中国计量大学浙江省智能制造质量大数据溯源与应用重点实验室开放课题(ZNZZSZCJLU2022-01);2021年国家级大学生创新创业训练计划项目(202110338013)资助。

摘　　要：睡枕的高度对睡眠质量有着决定性的影响,研究表明:为保证高质量的睡眠,人体侧卧时的枕头高度应大于平躺时的枕头高度。提出一种基于改进深度迁移学习的自适应睡枕设计方案:首先采用9路压力传感器阵列对肩颈部压力数据进行采集;然后,将9路压力信号转换为柱状分布图图像信号;接着提出一种改进的ResNet18+CBAM深度网络,并引入迁移学习对模型参数进行学习训练。实验结果表明:该方案具有硬件实现简单、识别准确率高的优点,可以满足睡枕产品开发的需求。The height of the pillow has a decisive impact on the quality of sleep,and studies have shown that the height of the pillow when lying on the side should be greater than the height of the pillow when lying flat.This paper presents an adaptive sleeping pillow design scheme based on improved deep transfer learning.Firstly,the 9-way pressure sensors are used to collect the pressure data of the shoulder and neck.Then,the 9-way pressure signals are converted into a histogram image signal.Next,an improved ResNet18+CBAM deep network is proposed,and transfer learning mechanism is introduced to train and recognize the sleeping posture.The experimental results show that the scheme has the advantages of simple hardware implementation and high recognition accuracy in software.

关 键 词：睡姿识别 残差神经网络 迁移学习 柱状分布图 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] R318[自动化与计算机技术—控制科学与工程]