Fall Detection With Multiple Cameras: An Occlusion-Resistant Method Based on 3-D Silhouette Vertical Distribution

Abstract

According to the demographic evolution in industrialized countries, more and more elderly people will experience falls at home and will require emergency services. The main problem comes from fall-prone elderly living alone at home. To resolve this lack of safety, we propose a new method to detect falls at home, based on a multiple-cameras network for reconstructing the 3-D shape of people. Fall events are detected by analyzing the volume distribution along the vertical axis, and an alarm is triggered when the major part of this distribution is abnormally near the floor during a predefined period of time, which implies that a person has fallen on the floor. This method was validated with videos of a healthy subject who performed 24 realistic scenarios showing 22 fall events and 24 cofounding events (11 crouching position, 9 sitting position, and 4 lying on a sofa position) under several camera configurations, and achieved 99.7% sensitivity and specificity or better with four cameras or more. A real-time implementation using a graphic processing unit (GPU) reached 10 frames per second (fps) with 8 cameras, and 16 fps with 3 cameras.

Vision-Based Fall Detection with Convolutional Neural Networks

“…Moreover, Auvinet et al [25] used a 3D vision system, which provides them with more information about the entire scenario than in our approach, although we are only three points behind in specificity from their results, showing that our system performs comparably even with less information.…”

“…The Kinect camera is very popular given its low price and high performance. Auvinet et al [25] used a Kinect camera to build a 3D silhouette to then analyze the volume distribution along the vertical axis. Gasparrini et al [26] also used such a camera to extract 3D features and then applied a tracking system to detect the falls.…”

“…(ii) Regarding the work by Auvinet et al [25] and Rougier et al [19], they are not included in the comparison tables due to their evaluation methodology (http://www.iro.umontreal.ca/∼labimage/Dataset/ technicalReport.pdf). They used a criterion that sets a variable called fall , representing the frame where the fall starts, therefore dividing each video into two parts.…”

“…Furthermore, Auvinet et al [25] used ad hoc threshold values to detect a lying-on-the-floor position, not the fall sequence itself. With this method, they reported sensitivity and specificity values up to 99.70%, while we obtained 99.00% and 96.00%, respectively.…”

Recognition of Falls and Activities of Daily Living using Tri-axial Accelerometer and Bi-axial Gyroscope

“…Since falls can have a devastating effect on the overall health and quality of living, many researches have actively been undertaken to monitor falls and activities of daily living (ADL) including walking patterns and postural changes. Several falls detection methods, using camera and video [2][3][4], tri-axial accelerometer [5,6], and gyroscope [7,8], have been studied to monitor falls. Among them, the tri-axial accelerometers and bi-axial gyroscope have been extensively used since these sensors have a good performance with high accuracy and reproducibility despite a relatively-low price.…”