Rainer Planinc

Introducing the use of depth data for fall detection

Current emergency systems for elderly contain at least one sensor (button or accelerometer), which has to be worn or pressed in case of emergency. If elderly fall and loose their consciousness, they are not able to press the button anymore. Therefore, autonomous systems to detect falls without wearing any devices are needed. This paper presents three different non-invasive technologies: the use of audio, 2D sensors (cameras) and introduces a new technology for fall detection: the Kinect as 3D depth sensor. Our fall detection algorithms using the Kinect are evaluated on 72 video sequences, containing 40 falls and 32 activities of daily living. The evaluation results are compared with State-of-the-Art approaches using 2D sensors or microphones.

Robust fall detection by combining 3d data and fuzzy logic

Falls are a major risk for the elderly and where immediate help is needed. The elderly, especially when suffering from dementia, are not able to react to emergency situations properly, thus falls need to be detected automatically. An overview of different classes of fall detection approaches is presented and a vision-based approach is introduced. We propose the use of a Kinect to obtain 3D data in combination with fuzzy logic for robust fall detection and show that our approach outperforms current state-of-the-art algorithms. Our approach is evaluated on 72 video sequences, containing 40 falls and 32 activities of daily living.

Emergency System for Elderly – A Computer Vision Based Approach

Elderly tend to forget or refuse wearing devices belonging to an emergency system (e.g. panic button). A vision based approach does not require any sensors to be worn by the elderly and is able to detect falls automatically. This paper gives an overview of my thesis, where different fall detection approaches are evaluated and combined. Furthermore, additional knowledge about the scene is incorporated to enhance the robustness of the system. To verify its feasibility, extensive tests under laboratory settings and real environments are conducted.

Fall Detection Based on Depth-Data in Practice

Falls are a leading cause of accidental deaths among the elderly population. The aim of fall detection is to ensure quick help for fall victims by automatically informing caretakers. We present a fall detection method based on depth-data that is able to detect falls reliably while having a low false alarm rate – not only under experimental conditions but also in practice. We emphasize person detection and tracking and utilize features that are invariant with respect to the sensor position, robust to partial occlusions, and computationally efficient. Our method operates in real-time on inexpensive hardware and enables fall detection systems that are unobtrusive, economic, and plug and play. We evaluate our method on an extensive dataset and demonstrate its capability under practical conditions in a long-term evaluation.

Detecting Changes in Elderly's Mobility Using Inactivity Profiles

Abnormal inactivity indicates situations, where elderly need assistance. Systems detecting the need for help models the amount of inactivity using inactivity profiles. Depending on the analysis of the profiles, events (e.g. falls) or long-term changes (decrease of mobility) are detected. Until now, inactivity profiles are only used to detect abnormal behavior on the short-term (e.g. fall, illness), but not on the long-term. Hence, this work introduces an approach to detect significant changes on mobility using long-term inactivity profiles, since these changes indicate enhanced or decreased mobility of elderly. Preliminary results are obtained by the analysis of the motion data of an elderly couple over the duration of 100 days and illustrates the feasibility of this approach.

Automated Timed Up & Go Test for functional decline assessment of older adults

Human mobility is an important health indicator, especially for older adults potentially transitioning to frailty. Currently, the analysis of human mobility is based on expensive or intrusive technologies. Depth camera devices, such as the Microsoft Kinect, have been demonstrated to be a valid low-cost alternative for assessing a persons mobility. In this work, mobility assessment is approached based on the automated analysis of the Timed Up & Go (TUG) test. Two methods based on depth and on skeleton data are proposed. In order to evaluate the proposed mobility analysis approaches, human mobility datasets have been acquired and manually labeled. It is shown that human mobility analysis based on off the shelf 3d sensors have the potential to assess functional decline of older adults.

Combining Technical and User Requirement Analysis to Support Wellbeing at the Workplace

The development of a technical system in order to support wellbeing of the workplace This work is supported by the EU and national funding organizations of EU member states under grant AAL 2013-6-063. demands for considering the requirements of the user, while developing state-of-the-art technology. Hence, in a first step, the requirements of the end user need to be analyzed as well as sensor technology of state-of-the-art sensors in order to match technology according to the users needs. Within this paper different sensors technologies are compared and the requirements of end user at the workplace are analyzed. By matching both, technological as well as sociological aspects allows for the development of technical system, fitting to the demands of end user.

Local behavior modeling based on long-term tracking data

Modeling the behavior of elderly people to detect changes in their health status or mobility is challenging and thus requires to combine temporal and spatial knowledge. Spatial knowledge is obtained by a novel human centered scene understanding approach, being able to accurately model sitting and walking regions based on noisy long-term tracking data from a depth sensor, without exploiting geometric information. A local behavior model based on the detected functional regions is introduced, allowing an in depth behavioral analysis. The proposed approaches are evaluated on three different datasets from two application domains (home and office environment), containing more than 180 days of tracking data.

Human Centered Scene Understanding Based on 3D Long-Term Tracking Data

Scene understanding approaches are mainly based on geometric information, not considering the behavior of humans. The proposed approach introduces a novel human-centric scene understanding approach, based on long-term tracking information. Long-term tracking information is filtered, clustered and areas offering meaningful functionalities for humans are modeled using a kernel density estimation. This approach allows to model walking and sitting areas within an indoor scene without considering any geometric information. Thus, it solely uses continuous and noisy tracking data, acquired from a 3D sensor, monitoring the scene from a bird’s eye view. The proposed approach is evaluated on three different datasets from two application domains (home and office environment), containing more than 180 days of tracking data.

Ergonomic-Monitoring of Office Workplaces Using Kinect

Prolonged sitting is an aggravating factor in low back and neck pain. Increased use of computers at workplaces could therefore cause health risks. This paper evaluates the application of the Microsoft Kinect in order to in-vestigate the ergonomics at the place of employment. The Kinect is a cheap device and commercially available which enables the user to record 3D data of the human body. Within this paper, guidelines for the ’ideal’ placement of the Kinect are provided in order to enhance the robustness of the skeleton recog-nition algorithm. An evaluation of 35 sequences (7 different positions in com-bination with 5 different sitting postures) showed that placing the Kinect sen-sor slantingly forward at an angle of 20° (in front of the subject) the joint rec-ognition rate achieved 89.62%. According to these results, the device should be positioned between 20° to 45° in order to robustly track a sitting person.