Early versus Late Fusion in a Multiple Camera Network for Fall Detection ∗

Falling and not being able to stand up is one of the major risks for elderly people who live alone. Camera based fall detectors represent one of the solutions to this problem. In this paper we compare two approaches for the detection of falls based on multiple cameras, the early fusion approach and the late fusion approach. In the early fusion approach, multiple camera views are combined to reconstruct the 3D voxel volume of the human. Based on semantic driven features fall detection is done on this 3D volume, whereas in the late fusion fall detection is done in 2D and each camera decides on its own, if a fall has occurred. These individual decisions are then combined into an overall decision. Fuzzy logic is both used to estimate confidence values for a fall/no fall in the single cameras as well as in the final voting step. We describe and evaluate both methods and give results on 73 video sequences.

[1]  A. Bourke,et al.  Fall detection - Principles and Methods , 2007, 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[2]  O. Wilder‐Smith,et al.  How dangerous are falls in old people at home? , 1981, British medical journal.

[3]  Rita Cucchiara,et al.  Probabilistic posture classification for Human-behavior analysis , 2005, IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans.

[4]  Jean Meunier,et al.  Fall Detection from Human Shape and Motion History Using Video Surveillance , 2007, 21st International Conference on Advanced Information Networking and Applications Workshops (AINAW'07).

[5]  C. Rougier,et al.  Monocular 3D Head Tracking to Detect Falls of Elderly People , 2006, 2006 International Conference of the IEEE Engineering in Medicine and Biology Society.

[6]  Chen Wu,et al.  Distributed Vision Networks for Human Pose Analysis , 2008 .

[7]  Chia-Wen Lin,et al.  Compressed-domain Fall Incident Detection for Intelligent Homecare , 2007, J. VLSI Signal Process..

[8]  Bart Vanrumste,et al.  How to detect human fall in video? An overview , 2009 .

[9]  James M. Keller,et al.  Linguistic summarization of video for fall detection using voxel person and fuzzy logic , 2009, Comput. Vis. Image Underst..

[10]  C. Dyer Volumetric Scene Reconstruction from Multiple Views , 2001 .

[11]  Zhihai He,et al.  Recognizing Falls from Silhouettes , 2006, 2006 International Conference of the IEEE Engineering in Medicine and Biology Society.

[12]  Pau-Choo Chung,et al.  FALL DETECTION USING MODULAR NEURAL NETWORKS WITH BACK-PROJECTED OPTICAL FLOW , 2007 .

[13]  Stephen J. McKenna,et al.  Activity summarisation and fall detection in a supportive home environment , 2004, ICPR 2004.

[14]  A. Enis Çetin,et al.  HMM Based Falling Person Detection Using Both Audio and Video , 2005, 2006 IEEE 14th Signal Processing and Communications Applications.

[15]  Yap-Peng Tan,et al.  Fall Incidents Detection for Intelligent Video Surveillance , 2005, 2005 5th International Conference on Information Communications & Signal Processing.