Development of a gait assessment device using deep convolutional neural network and pressure distribution insole / Ismael S. Ay-ay, John Isa O. Palacio, Dale M. Panganiban, Selwyne Christian E. Ponce, Lanz Brent Salmo
Material type: TextPublication details: City of Batac : MMSU, 2024Description: xxii, 167 leaves : 29 cmSummary: This study addresses the limitations of traditional gait assessment methods by developing a Raspberry Pi-driven device that employs computer vision and deep learning to provide objective gait cycle analysis. Integrating a Deep Convolutional Neural Network (DCNN) model for gait phase classification and MediaPipe for detecting hip, knee, and ankle joints to calculate their range of motion, the device also includes pressure distribution insoles for enhanced analysis. Evaluations revealed high user satisfaction (average scores of 4.49 for software and 4.69 for hardware) and impressive accuracy of 96% for the DCNN model in gait phase classification. Despite some delays in pressure distribution synchronization, the device demonstrated significant potential in reducing subjectivity in gait analysis, offering a sophisticated tool for clinical diagnostics and rehabilitation. Future recommendations include developing a custom pose estimation model, upgrading hardware, incorporating frontal view analysis, and integrating additional sensors for comprehensive biomechanical data.Item type | Home library | Shelving location | Call number | Status | Notes | Date due | Barcode |
---|---|---|---|---|---|---|---|
Thesis/Dissertation | MMSU Main Library | Theses and Dissertation Section | Available | Room Use Only | UTHESIS-6928 |
Browsing MMSU Main Library shelves, Shelving location: Theses and Dissertation Section Close shelf browser (Hides shelf browser)
UTHESIS (Bachelor of Science in Computer Engineering)
Bibliography: leaves 85-89
This study addresses the limitations of traditional gait assessment methods by developing a Raspberry Pi-driven device that employs computer vision and deep learning to provide objective gait cycle analysis. Integrating a Deep Convolutional Neural Network (DCNN) model for gait phase classification and MediaPipe for detecting hip, knee, and ankle joints to calculate their range of motion, the device also includes pressure distribution insoles for enhanced analysis. Evaluations revealed high user satisfaction (average scores of 4.49 for software and 4.69 for hardware) and impressive accuracy of 96% for the DCNN model in gait phase classification. Despite some delays in pressure distribution synchronization, the device demonstrated significant potential in reducing subjectivity in gait analysis, offering a sophisticated tool for clinical diagnostics and rehabilitation. Future recommendations include developing a custom pose estimation model, upgrading hardware, incorporating frontal view analysis, and integrating additional sensors for comprehensive biomechanical data.
There are no comments on this title.