Privacy-Preserving Pose Detection using Multi Low-Resolution Infrared Array Sensors
We are confronted with the social problem that the number of nursing staff is decreasing while the number of people in need of care is increasing. Especially in the case of home intensive care patients, care is particularly time-consuming and a high degree of concentration is required from the caregivers because even one mistake can have fatal consequences. The relief of such safety critical scenarios is important for the health of all parties involved. Providing the exact localization of the patient’s whereabouts and pose can be advantageous in various scenarios, for example in the case of long-term home-ventilated patients, the presence of a nurse must be ensured during cleaning or maintenance work on the ventilator, or the recognition of the posture is necessary for the correct behaviour of robotic assistance systems. Despite the fact that localization is an important factor, the privacy of patient and caregiver should not be invaded, especially in the home environment. The usage of cameras for 2D or 3D image data recording is usually rejected. In order to preserve the well-being of the patients and provide the necessary information for the assistance systems at the same time, the present research work deals with the possibility to determine the current position in space as well as the posture by using several infrared array sensors. As these sensors register only low-resolution temperature matrices, no conclusions about the patient’s identity or activities including minor movements can be drawn. For the pose detection, an attempt is made to infer the states "standing", "sitting" and "lying". For this particular purpose, sensor clusters consisting of three sensors each are mounted on the wall and on the ceiling in order to capture an overall picture of the scenario. The data captured is used to train a neural network to determine the position and posture of the captured person. The results of the work carried out show that the data of the low-resolution sensor clusters are sufficient for the determination and can therefore preserve the privacy of the patient despite the provision of the necessary information.
Literatur- Rapoport, M. (2012). The home under surveillance: A tripartite
assemblage. Surveillance & Society, 10(3/4):320.
- Gerka, A., Pfingsthorn, M., Lupkes, C., Sparenberg, K.,
Frenken, M., Lins, C., and Hein, A. (2018). Detecting the
number of persons in the bed area to enhance the safety
of artificially ventilated persons. In 2018 IEEE 20th International
Conference on e-Health Networking, Applications
and Services (Healthcom), pages 1–6. IEEE.
- Guler, A., Kardaris, N., Chandra, S., Pitsikalis, V., Werner, C., Hauer, K., ... & Kokkinos, I. (2016, October). Human joint angle estimation and gesture recognition for assistive robotic vision. In European Conference on Computer Vision (pp. 415-431). Springer, Cham.
Im Artikelarchiv von hpsmedia nach Christian Kowalski suchen