Attention-Based Spatial Temporal Fusion Deep Learning Sleeping Posture Monitoring Model
SHI Yongwu1, LI Xiaoyong2, SHI Yongde3, SHI Yongmin4, XIE Quan5
1. Department of Equipment, Guizhou Provincial People’s Hospital, Guiyang Guizhou 550002, China; 2. School of Environment,
South China Normal University, Guangzhou Guangdong 51000, China; 3. Health Center in Dashan Town, Panzhou Guizhou 553507,
China; 4. Urban Planning and Construction Management Institute of Dashan Town, Panzhou Guizhou 553507, China;
5. College of Big Data and Information Engineering, Guizhou University, Guiyang Guizhou 550025, China
Abstract: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.
石用伍1,李小勇2,石用德3,石用民4,谢泉5. 基于注意力机制的空时融合深度学习睡姿监测算法研究[J]. 中国医疗设备, 2022, 37(7): 39-44.
SHI Yongwu1, LI Xiaoyong2, SHI Yongde3, SHI Yongmin4, XIE Quan5. Attention-Based Spatial Temporal Fusion Deep Learning Sleeping Posture Monitoring Model. China Medical Devices, 2022, 37(7): 39-44.