Sleep problems affect millions of people globally, presenting a tremendous burden on society and people. Sleep monitoring is important to understand patient-specific problems and to guide treatment. However, the traditional methods of sleep quality detection are inconvenient, costly, and time-consuming. In this research, we conducted a user study to investigate if sensors built into smartphones can be utilized for passive sleep duration and quality monitoring. Smartphone sensor and ‘ActiGraph’ ground truth data were simultaneously collected from 6 subjects. The data were preprocessed data to remove errors and inconsistencies. The preprocessed data was extracted into seven types of features that were then used to train machine learning classification models. In rigorous evaluation, gradient boosting machine learning algorithm performed best, achieving an accuracy of 88.08%. The final model was used to create a tool that will extract and analyze patient’s sleep features from smartphone sensors data. In the future, this model can be used to develop a smartphone App to monitor users’ sleep quality automatically.

Creator

• Peng, Yuxiao

Mitwirkende

• Agu, Emmanuel O.
• Korkin, Dmitry

Degree

• MS

Unit

• Bioinformatics & Computational Biology

Publisher

• Worcester Polytechnic Institute

Identifier

• etd-3966

Advisor

• Agu, Emmanuel O.

Defense date

• 2021-05-13

Year

• 2020

Date created

• 2020-05-18

Resource type

• Thesis

Rights statement

• In Copyright

Zuletzt geändert

• 2023-10-06