Vision Transformers (ViT) for Blanket-Penetrating Sleep Posture Recognition Using a Triple Ultra-Wideband (UWB) Radar System

Abstract

Sleep posture has a crucial impact on the incidence and severity of obstructive sleep apnea (OSA). Therefore, the surveillance and recognition of sleep postures could facilitate the assessment of OSA. The existing contact-based systems might interfere with sleeping, while camera-based systems introduce privacy concerns. Radar-based systems might overcome these challenges, especially when individuals are covered with blankets. The aim of this research is to develop a nonobstructive multiple ultra-wideband radar sleep posture recognition system based on machine learning models. We evaluated three single-radar configurations (top, side, and head), three dual-radar configurations (top + side, top + head, and side + head), and one tri-radar configuration (top + side + head), in addition to machine learning models, including CNN-based networks (ResNet50, DenseNet121, and EfficientNetV2) and vision transformer-based networks (traditional vision transformer and Swin Transformer V2). Thirty participants (n = 30) were invited to perform four recumbent postures (supine, left side-lying, right side-lying, and prone). Data from eighteen participants were randomly chosen for model training, another six participants’ data (n = 6) for model validation, and the remaining six participants’ data (n = 6) for model testing. The Swin Transformer with side and head radar configuration achieved the highest prediction accuracy (0.808). Future research may consider the application of the synthetic aperture radar technique.

Research on Diesel Engine Fault Status Identification Method Based on Synchro Squeezing S-Transform and Vision Transformer

“…The Transformer model is based entirely on a self-attentive mechanism without any convolutional or recurrent neural network layers and is not subject to local interaction limitations [ 25 ]. ViT was the first Transformer model used to replace CNNs and applied to image classification [ 26 ]. Although Transformer was originally applied to sequence-to-sequence learning on text data, it has now been extended to various modern deep learning in areas such as vision, target detection and image segmentation [ 27 ].…”

A smart detection method for sleep posture based on a flexible sleep monitoring belt and vital sign signals

“…Moreover, the installation of each camera is very complex since it should be aimed at the bed and not obstructed [ [22] , [23] , [24] , [25] , [26] , [27] , [28] , [29] ]. Radio frequency (RF) sensors, such as single-tone continuous Frequency Modulated Continuous Wave (FMCW), ultra-wideband (UWB), and millimeter-wave radar sensors, are also suitable for monitoring sleep posture and vital signs [ [30] , [31] , [32] ]. These sensors generally offer high detection accuracy but can be susceptible to interference when multiple people are present in the same area.…”

“…For vital signs based classifier, vital sign signals acquired by MEMS IMU, ElectroCardioGraphy (ECG), BCG, SCG or GCG device are fed into this classifier to identify the sleep posture. After pre-processing of the vital sign signals, K-nearest neighbor (KNN), Naive Bayes (NB), Decision Tree (DT), ExtraTree (ET), K-means clustering, Swin Transformer (ST), SVM, CNN are adopted for feature extraction and sleep posture recognition, and the detection accuracy ranges from 80.8 % to 99.67 % [ 32 , [42] , [43] , [44] , [45] , [46] ]. The detection accuracies of some vital signs based classifiers are very high, especially for machine learning models.…”

A Deep Learning Method of Human Identification from Radar Signal for Daily Sleep Health Monitoring

“…Various studies have chosen radar signals as a non-contact source for vital biometric extraction, motion monitoring, and pose estimation [5][6][7][8][9][10][11], which suggests that, from radar data, we can acquire highly personal information, which is potentially promising for human identification. Some groups [12][13][14] have shown that with machine learning techniques, the features of motions such as walking and running can be promising for trajectory identification and tracking.…”