基于注意力机制的空时融合深度学习睡姿监测算法研究  

Attention-Based Spatial Temporal Fusion Deep Learning Sleeping Posture Monitoring Model

作　　者：石用伍[1] 李小勇 石用德 石用民 谢泉 SHI Yongwu;LI Xiaoyong;SHI Yongde;SHI Yongmin;XIE Quan(Department of Equipment,Guizhou Provincial People’s Hospital,Guiyang Guizhou 550002,China;School of Environment,South China Normal University,Guangzhou Guangdong 51000,China;Health Center in Dashan Town,Panzhou Guizhou 553507,China;Urban Planning and Construction Management Institute of Dashan Town,Panzhou Guizhou 553507,China;College of Big Data and Information Engineering,Guizhou University,Guiyang Guizhou 550025,China)

机构地区：[1]贵州省人民医院设备处,贵州贵阳550002 [2]华南师范大学环境学院,广东广州51000 [3]贵州省盘州市大山镇卫生院,贵州盘州553507 [4]贵州省盘州市大山镇城镇规划建设管理所,贵州盘州553507 [5]贵州大学大数据与信息工程学院,贵州贵阳550025

出　　处：《中国医疗设备》2022年第7期39-44,共6页China Medical Devices

基　　金：国家自然科学基金(61264004);广东省企业科技特派员项目(GDKTP2020031800)。

摘　　要：目的针对用于无扰睡姿检测的心冲击图(Ballistocardiogram,BCG)信号特征微弱,并且具有非线性、非平稳性强、存在噪声干扰以及信号本身具有空间和时域信息的特点,提出了一种基于注意力机制融合空时特征的深度学习睡姿检测模型(Deep Learning Sleeping Posture Monitoring Model,CTAM)。方法CTAM是一种可实现端到端的实时睡姿检测方案,通过睡眠带测试真实睡眠状态下睡姿的BCG信号,并构建数据集进行仿真对比实验。结果与具有类似结构的传统卷积神经网络(Convolutional Neural Networks,CNN)模型和空时融合的卷积-长短时记忆网络(Convolutional Long Short-Term Memory Network,CNN-LSTM)相比,CTAM在训练集的收敛性和测试集的准确率上均有显著的提升,其中,在测试集上的准确率分别较CNN模型和CNN-LSTM模型提升了1.46%和4.61%。结论CTAM这种算法模型在基于BCG信号下能实现睡姿的实时、有效、无扰监测,在改善睡眠质量监测领域具有较好的应用前景。Objective This paper focuses on the problems of ballistocardiogram(BCG)signal used for undisturbed sleep position detection has weak signal characteristics,which are non-linearity,strong non-stationarity,noise interference,and the signal itself has spatial and temporal information.A deep learning sleeping posture monitoring model(CTAM)based on attention mechanism and spatial features was proposed.Methods CTAM is an end-to-end real-time sleeping posture detection scheme.BCG signal of sleeping posture in real sleep is tested through the sleep belt,and a data set is constructed for simulation and comparison experiments.Results The results showed that compared with the traditional convolutional neural networks(CNN)model with similar structure and the space-time fusion convolutional-long short-term memory network(CNN-LSTM),CTAM significantly improved the convergence of training set and the accuracy of test set,and the accuracy of test set was 1.46%and 4.61%higher than CNN model and CNN-LSTM model,respectively.Conclusion CTMA algorithm model based on the BCG signal to achieve real-time sleeping position,effective and non-disturbed monitoringof sleeping position,which has a good application prospect in the field of improving sleep quality monitoring.

关 键 词：睡姿 心冲击图 卷积神经网络 深度学习睡姿检测模型 注意力机制 

分 类 号：R318[医药卫生—生物医学工程]