Publication: Sleep posture recognition for bedridden patient
Issued Date
2019-01-01
Resource Type
ISSN
18761119
18761100
18761100
Other identifier(s)
2-s2.0-85051140763
Rights
Mahidol University
Rights Holder(s)
SCOPUS
Bibliographic Citation
Lecture Notes in Electrical Engineering. Vol.513, (2019), 79-87
Suggested Citation
Nitikorn Srisrisawang, Lalita Narupiyakul Sleep posture recognition for bedridden patient. Lecture Notes in Electrical Engineering. Vol.513, (2019), 79-87. doi:10.1007/978-981-13-1059-1_8 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/50865
Research Projects
Organizational Units
Authors
Journal Issue
Thesis
Title
Sleep posture recognition for bedridden patient
Author(s)
Other Contributor(s)
Abstract
© Springer Nature Singapore Pte Ltd. 2019. One type of patients that needs to live on the bed for a certain time or worst, for the rest of their life is called bedridden. This type of patients need special attention from caretaker to regularly change the posture of the patient in order to prevent symptom named bed sore or pressure sore which will happen when the weight of the patient is applied to some points of the body too long which leads to injury to that certain points of the body. This research will carried out to design a system to relieve the work for the caretaker of a bedridden patient. This system consists of three parts; Sleep data collection where computer that connected to Kinect will continuously monitor the patient and send the data to the next part, Sleep posture analysis which will determine the postures of the patient from the input data, and Sleep notification part which will notify user with the current state of the patient. There are 3 machine learning algorithms that were chosen to compare their performance; Decision Tree (DT), Neural Network (NN), and Support Vector Machine (SVM). In the case of using the data from the same subjects as in the training set, DT shows lower accuracy at 93.33% than NN and SVM which achieve 100%. Similarly, in the case of using dataset that is not in the training set, DT still performs at 90% while both NN and SVM achieve 100%, the data are tested from both the subjects within the training set and new subjects but without any error exclusion which illustrates that NN which achieves 63.33% accuracy is more prone to the data with error than SVM which is 57.78%. Hence, NN is implemented with the system.