Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorAmini, A-
dc.contributor.authorBanitsas, K-
dc.contributor.authorCosmas, J-
dc.identifier.citation2016 IEEE International Symposium on Medical Measurements and Applications, MeMeA 2016 - Proceedings, (2016)en_US
dc.description.abstractIn this paper, two algorithms were tested on 11 healthy adults: one based on heuristic and another one on video tagging machine learning methods for automatic fall detection; both utilizing Microsoft Kinect v2. For our heuristic approach, we used skeletal data to detect falls based on a set of instructions and signal filtering methods. For the machine learning approach, we implemented a dataset utilizing the Adaptive Boosting Trigger (AdaBoostTrigger) algorithm via video tagging to enable fall detection. For each approach, each subject on average has performed six true positive and six false positive fall incidents in two different conditions: one with objects partially blocking the sensor's view and one with partial obstructed field of view. The accuracy of each approach has been compared against one another in different conditions. The result showed an average of 95.42 % accuracy in the heuristic approach and 88.33 % in machine learning technique. We conclude that heuristic approach performs more accurately for fall detection when there is a limited number of training dataset available. Nevertheless, as the gesture detection's complexity increases, the need for a machine learning technique is inevitable.en_US
dc.subjectFall detectionen_US
dc.subjectMachine learningen_US
dc.titleA comparison between heuristic and machine learning techniques in fall detection using Kinect v2en_US
dc.typeConference Paperen_US
dc.relation.isPartOf2016 IEEE International Symposium on Medical Measurements and Applications, MeMeA 2016 - Proceedings-
Appears in Collections:Dept of Electronic and Electrical Engineering Research Papers

Files in This Item:
File Description SizeFormat 
Fulltext.pdf1.26 MBAdobe PDFView/Open

Items in BURA are protected by copyright, with all rights reserved, unless otherwise indicated.