Deep Residual Networks for Sleep Posture Recognition With Unobtrusive Miniature Scale Smart Mat System

Diao, Haikang; Chen, Chen; Yuan, Wei; Amara, Amara; Tamura, Toshiyo; Fan, Jiahao; Meng, Long; Liu, Xiangyu; Chen, Wei

doi:10.1109/tbcas.2021.3053602

Cited by 38 publications

(21 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With reference to laying posture recognition, the proposed FCN obtains an average accuracy value of 96.77% to classify 17 laying postures in real-time exploiting 108 sensors, and a throughput of 9.13 kHz. The state-of-the-art in this case is represented by the recent work in [36], dealing with in-bed posture recognition, exploiting a microcontroller unit to implement a very complex ResNet composed by 17 CONV layers, two MaxPool and three FC layers, in order to obtain an average classification accuracy of 95.08% by using 1024 force sensitive resistor sensors.…”

Section: Comparison With the Literaturementioning

confidence: 99%

A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems

et al. 2021

View full text Add to dashboard Cite

A custom HW design of a Fully Convolutional Neural Network (FCN) is presented in this paper to implement an embeddable Human Posture Recognition (HPR) system capable of very high accuracy both for laying and sitting posture recognition. The FCN exploits a new base-2 quantization scheme for weight and binarized activations to meet the optimal trade-off between low power dissipation, a very reduced set of instantiated physical resources and state-of-the-art accuracy to classify human postures. By using a limited number of pressure sensors only, the optimized HW implementation allows keeping the computation close to the data sources according to the edge computing paradigm and enables the design of embedded HP systems. The FCN can be simply reconfigured to be used for laying and sitting posture recognition. Tested on a public dataset for in-bed posture classification, the proposed FCN obtains a mean accuracy value of 96.77% to recognize 17 different postures, while a small custom dataset has been used for training and testing for sitting posture recognition, where the FCN achieves 98.88% accuracy to recognize eight positions. The FCN has been prototyped on a Xilinx Artix 7 FPGA where it exhibits a dynamic power dissipation lower than 11 mW and 7 mW for laying and sitting posture recognition, respectively, and a maximum operation frequency of 47.64 MHz and 26.6 MHz, corresponding to an Output Data Rate (ODR) of the sensors of 16.50 kHz and 9.13 kHz, respectively. Furthermore, synthesis results with a CMOS 130 nm technology have been reported, to give an estimation about the possibility of an in-sensor circuital implementation.

show abstract

Section: Comparison With the Literaturementioning

confidence: 99%

A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems

et al. 2021

View full text Add to dashboard Cite

show abstract

“…In the study of sEMG signal characteristics ( Feng et al, 2011 ; Liu et al, 2013 ), the feature extraction of sEMG signals is mainly based on manual methods. In this paper, the dimensional-reduction signals are directly fed into the DRSN network, and the features irrelevant to the classification task are zeroed by soft thresholding through the attention mechanism ( Diao et al, 2021 ). The detailed process of the DRSN algorithm is as follows:…”

Section: Methodsmentioning

confidence: 99%

A comparison of contributions of individual muscle and combination muscles to interaction force prediction using KPCA-DRSN model

Lü

Gao

Cao

et al. 2022

Front. Bioeng. Biotechnol.

View full text Add to dashboard Cite

Rapid and accurate prediction of interaction force is an effective way to enhance the compliant control performance. However, whether individual muscles or a combination of muscles is more suitable for interaction force prediction under different contraction tasks is of great importance in the compliant control of the wearable assisted robot. In this article, a novel algorithm that is based on sEMG and KPCA-DRSN is proposed to explore the relationship between interaction force prediction and sEMG signals. Furthermore, the contribution of each muscle to the interaction force is assessed based on the predicted results. First of all, the experimental platform for obtaining the sEMG is described. Then, the raw sEMG signal of different muscles is collected from the upper arm during different contractions. Meanwhile, the output force is collected by the force sensor. The Kernel Principal Component Analysis (KPCA) method is adopted to remove the invalid components of the raw sEMG signal. After that, the processed sequence is fed into the Deep Residual Shrinkage Network (DRSN) to predict the interaction force. Finally, based on the prediction results, the contribution of each sEMG signal from different muscles to the interaction force is evaluated by the mean impact value (MIV) indicator. The experimental results demonstrate that our methods can automatically extract the valid features of sEMG signal and provided fast and efficient prediction. In addition, the single muscle with the largest MIV index could predict the interaction force faster and more accurately than the muscle combination in different contraction tasks. The finding of our research provides a solid evidence base for the compliant control of the wearable robot.

show abstract

“…The sleep posture classification has taken its advancement in sleep posture recognition with a matching-based approach termed as Body-Earth mover's distance. The similarity level of a posture is identified using weighted 2-dimensional shapes combined with the earth mover's distance and Euclidean distance for estimation of the similarity index [ 13 ]. Neither multiple features nor unobtrusive sleep posture is collected to recognize the sleep posture.…”

Section: Literature Surveymentioning

confidence: 99%

“…Unlike the above two [ 13 ] preprocessing procedures used to normalize the image dataset from MPII and the dataset captured from the subjects considered for the experiment, the heartbeat rate corresponding to the movement of the human body adopts principal component analysis that will rotate the accelerometer sensor in order to capture the variations during sleep. The angle of rotation is given by the leading eigenvector of the PCA and the protracted edge of the sleep image.…”

Section: System Design and Implementationmentioning

confidence: 99%

An Enhanced Posture Prediction-Bayesian Network Algorithm for Sleep Posture Recognition in Wireless Body Area Networks

Roshini

Kiran

2022

International Journal of Telemedicine and Applications

View full text Add to dashboard Cite

Wireless body area networks have taken their unique recognition in providing consistent facilities in health monitoring. Several studies influence physiological signal monitoring through a centralized approach using star topology in regular activities like standing, walking, sitting, and running which are considered active postures. Unlike regular activities like walking, standing, sitting, and running, the in-bed sleep posture monitoring of a subject is highly necessary for those who have undergone surgery, victims of breathing problems, and victims of COVID-19 for whom oxygen imbalance is a major issue as the mortality rate in sleep is high due to unattended patients. Suggestions from the medical field state that the patients with the above-mentioned issues are highly suggested to follow the prone sleep posture that enables them to maintain the oxygen level in the human body. A distributed model of communication is used where mesh topology is used for the data packets to be carried in a relay fashion to the sink. Heartbeat rate (HBR) and image monitoring of the subject during sleep are closely monitored and taken as input to the proposed posture prediction-Bayesian network (PP-BN) to predict the consecutive postures to increase the accuracy rate of posture recognition. The accuracy rate of the model outperforms the existing classification and prediction algorithms which take the cleaned dataset as input for better prediction results.

show abstract

Deep Residual Networks for Sleep Posture Recognition With Unobtrusive Miniature Scale Smart Mat System

Cited by 38 publications

References 24 publications

A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems

A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems

A comparison of contributions of individual muscle and combination muscles to interaction force prediction using KPCA-DRSN model

An Enhanced Posture Prediction-Bayesian Network Algorithm for Sleep Posture Recognition in Wireless Body Area Networks

Contact Info

Product

Resources

About