Smart Homes: An occlusion-resistant fall detection system for the elderly
2022 (engelsk)Independent thesis Advanced level (degree of Master (Two Years)), 80 poäng / 120 hp
Oppgave
Abstract [en]
Falls represent one of the most severe issues faced by the elder population. Therefore,in this work, we propose a system for fall detection using RGB camera images. In thisregard, we propose, associated to multiple classifiers, a fusion between features calculatedfrom the head region and those calculated from the whole body’s region. These regionsare extracted using the YOLOv5 object detector. Experiments conducted on a standardfall dataset showed satisfying and promising results. The system based on features fromthe head data yielded an accuracy of 94.79%, while the system based on fusion betweenhead and body features produced an accuracy of 98.08%.
sted, utgiver, år, opplag, sider
2022. , s. 62
Emneord [en]
fall detection, YOLOv5, SVM, K-NN, RFC, MLP
HSV kategori
Identifikatorer
URN: urn:nbn:se:ltu:diva-92684OAI: oai:DiVA.org:ltu-92684DiVA, id: diva2:1690722
Fag / kurs
Student thesis, at least 30 credits
Utdanningsprogram
Master Programme in Green Networking and Cloud Computing
Presentation
2022-06-08, Amphi 7, Faculté des Sciences & Technologies, Nancy, France, 15:30 (engelsk)
Veileder
Examiner
2022-11-102022-08-262022-11-10bibliografisk kontrollert