Design and Architecture of Posture Monitoring Belt equipped with Sensors for sitting persons
Presented by
Ferdews TLILI
On December 14th , 2024 at 11H00 AM, in Amphitheater Ibn Kholdoun, SUP'COM 2.
Thesis Committee
|
President |
Pr. Mourad MENIF |
SUP'COM, Tunisia |
|
Reviewers |
Pr. Taher EZZEDINE |
ENIT, Tunisia |
|
|
Pr. Faouzi ZARAI |
ENET’COM, Sfax, Tunisia |
|
Examiner |
Pr. Kaouthar SETHOM |
ENI Carthage, Tunisia |
|
Supervisor |
Pr. Ridha BOUALLEGUE |
SUP'COM, Tunisia |
|
Co-supervisor |
Dr. Rim HADDAD |
LAVAL University, Canada |
|
|
|
|
Spinal pain caused by the bad sitting posture can affect adults and young people. The slouching and leaning for long hours using computers and portable electronic devices are the main causes of the back problems. Recently, in order to prevent the spinal pain, sitting posture monitoring systems have been developed in the literature to assess the posture of a seated person in real-time and improve sitting posture. Camera-based systems are widely used to monitor sitting posture. These systems are based on image processing which can lead to an invasion of user’s privacy. In addition, the accuracy and reliability of these systems depend on the set-up implementation and positioning of the cameras. With the increasing variety of information provided by sensors are being integrated into the sitting posture monitoring system to overcome the limitations of camera-based systems. In this thesis, we analyzed existing systems in the literature. We then proposed a new architecture for a sitting posture monitoring system based on inertial sensors. In this proposal, we introduce shoulder flexion monitoring in addition to trunk flexion monitoring to provide comprehensive information on back posture compared with existing systems that are limited to trunk flexion. Furthermore, we have proposed an optimal sensor position with minimum sensors implemented without impacting on system functionality. The proposed posture monitoring system is a smart belt equipped with inertial sensors and smartphone applications to warn the person in case of detection of bad posture. We continued to improve the effectiveness of the proposed system by investigating the application of machine learning to our proposed system. We concluded by introducing the Random Forest with 30 trees to predict posture, as it presents the best accuracy and appropriate processing time for the real- time system compared with other machine learning algorithms applicable to the proposal.
Maintenant, allez pousser vos propres limites et réussir!