Research Article
Edge-Computing System Based on Smart Mat for Sleep Posture Recognition in IoMT
@INPROCEEDINGS{10.1007/978-3-031-06368-8_6, author={Haikang Diao and Chen Chen and Xiangyu Liu and Amara Amara and Wei Chen}, title={Edge-Computing System Based on Smart Mat for Sleep Posture Recognition in IoMT}, proceedings={Wireless Mobile Communication and Healthcare. 10th EAI International Conference, MobiHealth 2021, Virtual Event, November 13--14, 2021, Proceedings}, proceedings_a={MOBIHEALTH}, year={2022}, month={6}, keywords={Edge computing Sleep posture recognition EdgeNet Model quantization}, doi={10.1007/978-3-031-06368-8_6} }- Haikang Diao
Chen Chen
Xiangyu Liu
Amara Amara
Wei Chen
Year: 2022
Edge-Computing System Based on Smart Mat for Sleep Posture Recognition in IoMT
MOBIHEALTH
Springer
DOI: 10.1007/978-3-031-06368-8_6
Abstract
Sleep posture has been proven to be a crucial index for sleep monitoring in the Internet of Medical Things (IoMT). In this paper, an edge-computing system based on a smart mat for sleep posture recognition in IoMT is proposed. The system can recognize postures unobtrusively with a dense flexible sensor array. To meet the requirements of embedded system in IoMT, a light-weight algorithm that includes pre-processing, EdgeNet pre-training, model quantization, model deployment is proposed. Finally, the complete algorithm is deployed in embedded systems (STM32) and edge computing for sleep posture monitoring is implement in IoMT. Through a series of short-term and overnight experiments with 21 subjects, results exhibit that the accuracy of the short-term experiment is up to 92.10% and the overnight experiment is up to 75.43%. After quantization, the accuracy of the overnight is up to 74.79%, and the runtime of the complete algorithm is about 65ms in the STM32. Compared with other methods, edge-computing systems have the advantages of low power consumption, low cost, low latency, high reliability, and no risk of privacy leakage. With the promising results, the proposed system is capable of providing sleep posture recognition and can be integrated into IoMT as an edge device.
