Edouard Auvinet

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

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.

Fall detection from depth map video sequences

Falls are one of the major risks for seniors living alone at home. Computer vision systems, which do not require to wear sensors, offer a new and promising solution for fall detection. In this work, an occlusion robust method is presented based on two features: human centroid height relative to the ground and body velocity. Indeed, the first feature is an efficient solution to detect falls as the vast majority of falls ends on the ground or near the ground. However, this method can fail if the end of the fall is completely occluded behind furniture. Fortunately, these cases can be managed by using the 3D person velocity computed just before the occlusion.

Pose estimation with a Kinect for ergonomic studies: evaluation of the accuracy using a virtual mannequin.

Analyzing human poses with a Kinect is a promising method to evaluate potentials risks of musculoskeletal disorders at workstations. In ecological situations, complex 3D poses and constraints imposed by the environment make it difficult to obtain reliable kinematic information. Thus, being able to predict the potential accuracy of the measurement for such complex 3D poses and sensor placements is challenging in classical experimental setups. To tackle this problem, we propose a new evaluation method based on a virtual mannequin. In this study, we apply this method to the evaluation of joint positions (shoulder, elbow, and wrist), joint angles (shoulder and elbow), and the corresponding RULA (a popular ergonomics assessment grid) upper-limb score for a large set of poses and sensor placements. Thanks to this evaluation method, more than 500,000 configurations have been automatically tested, which would be almost impossible to evaluate with classical protocols. The results show that the kinematic information obtained by the Kinect software is generally accurate enough to fill in ergonomic assessment grids. However inaccuracy strongly increases for some specific poses and sensor positions. Using this evaluation method enabled us to report configurations that could lead to these high inaccuracies. As a supplementary material, we provide a software tool to help designers to evaluate the expected accuracy of this sensor for a set of upper-limb configurations. Results obtained with the virtual mannequin are in accordance with those obtained from a real subject for a limited set of poses and sensor placements.

Head detection using Kinect camera and its application to fall detection

This article proposes a head detection algorithm for depth video provided by a Kinect camera and its application to fall detection. The proposed algorithm first detects possible head positions and then based on these positions, recognizes people by detecting the head and the shoulders. Searching for head positions is rapid because we only look for the head contour on the human outer contour. The human recognition is a modification of HOG (Histogram of Oriented Gradient) for the head and the shoulders. Compared with the original HOG, our algorithm is more robust to human articulation and back bending. The fall detection algorithm is based on the speed of the head and the body centroid and their distance to the ground. By using both the body centroid and the head, our algorithm is less affected by the centroid fluctuation. Besides, we also present a simple but effective method to verify the distance from the ground to the head and the centroid.

Fall detection using multiple cameras

Today, different ways are suggested to help elderly people in case of emergency. Our aim here is to propose a novel method, without any wearable device, to detect falls on the floor with a multiple cameras system. This proposal uses image analysis to localise people and reconstruct their 3D shape and position. The particularity of this contribution is the use of cameras sharing a large common field of view. Experimental results obtained with 14 different fall scenarios and 14 normal daily activities showed a 100% fall detection efficiency.

Multiple depth cameras calibration and body volume reconstruction for gait analysis

In the last decade, gait analysis has become one of the most active research topics in biomedical research engineering partly due to recent development of sensors and signal processing devices and more recently depth cameras. The latters can provide real-time distance measurements of moving objects. In this context, we present a new way to reconstruct body volume in motion using multiple active cameras from the depth maps they provide. A first contribution of this paper is a new and simple external camera calibration method based on several plane intersections observed with a low-cost depth camera which is experimentally validated. A second contribution consists in a body volume reconstruction method based on visual hull that is adapted and enhanced with the use of depth information. Preliminary results based on simulations are presented and compared with classical visual hull reconstruction. These results show that as little as three low-cost depth cameras can recover a more accurate 3D body shape than twenty regular cameras.

New Lower-Limb Gait Asymmetry Indices Based on a Depth Camera

Background: Various asymmetry indices have been proposed to compare the spatiotemporal, kinematic and kinetic parameters of lower limbs during the gait cycle. However, these indices rely on gait measurement systems that are costly and generally require manual examination, calibration procedures and the precise placement of sensors/markers on the body of the patient. Methods: To overcome these issues, this paper proposes a new asymmetry index, which uses an inexpensive, easy-to-use and markerless depth camera (Microsoft Kinect™) output. This asymmetry index directly uses depth images provided by the Kinect™ without requiring joint localization. It is based on the longitudinal spatial difference between lower-limb movements during the gait cycle. To evaluate the relevance of this index, fifteen healthy subjects were tested on a treadmill walking normally and then via an artificially-induced gait asymmetry with a thick sole placed under one shoe. The gait movement was simultaneously recorded using a Kinect™ placed in front of the subject and a motion capture system. Results: The proposed longitudinal index distinguished asymmetrical gait (p < 0.001), while other symmetry indices based on spatiotemporal gait parameters failed using such Kinect™ skeleton measurements. Moreover, the correlation coefficient between this index measured by Kinect™ and the ground truth of this index measured by motion capture is 0.968. Conclusion: This gait asymmetry index measured with a Kinect™ is low cost, easy to use and is a promising development for clinical gait analysis.

Lower limb movement asymmetry measurement with a depth camera

The gait movement seems simple at first glance, but in reality it is a very complex neural and biomechanical process. In particular, if a person is affected by a disease or an injury, the gait may be modified. The left-right asymmetry of this movement can be related to neurological diseases, segment length differences or joint deficiencies. This paper proposes a novel method to analyze the asymmetry of lower limb movement which aims to be usable in daily clinical practice. This is done by recording the subject walking on a treadmill with a depth camera and then assessing left-right depth differences for the lower limbs during the gait cycle using horizontal flipping and registration of the depth images half a gait cycle apart. Validation on 20 subjects for normal gait and simulated pathologies (with a 5 cm sole), showed that this system is able to distinguish the asymmetry introduced. The major interest of this method is the low cost of the material needed and its easy setup in a clinical environment.

Depth energy image for gait symmetry quantification

This paper introduces a new quantification method for gait symmetry based on depth information acquired from a structured light system. First, the new concept of Depth Energy Image is introduced to better visualize gait asymmetry problems. Then a simple index is computed from this map to quantify motion symmetry. Results are presented for six subjects with and without gait problems. Since the method is markerless and cheap, it could be a very promising solution in the future for gait clinics.

Gait analysis with multiple depth cameras

The gait movement seems simple at first glance, but in reality it is a very complex neural and biomechanical process. In particular, if a person is affected by a disease or an injury, the gait may be modified. To help detecting such change, we propose a new method based on multiple depth cameras. The aim of this paper is to show the possibility to reconstruct the body 3D volume in real time during gait in order to detect a pathological problem related to this movement and eventually improve diagnosis. Preliminary results showed that the system is sensitive to gait change produced by a heel prosthesis (heel cup) inserted in one shoe of subjects walking on a treadmill. The system detected a difference between maximal forward and backward positions of lower limbs for this pathological walk, a difference that was negligible for normal walk. These promising results were obtained with only 3 low cost depth cameras; we therefore believe that such methodology opens a new and affordable way for 3D volumetric gait analysis.