Posture and Motion Detection Using Acceleration Data for Context Aware Sensing in Personal Healthcare Systems

摘要

Detection of human motions is of great relevance especially in the field of telemedicine. The analysis of physiological information gains important, incorruptible additional information about the context in terms of patients' activities by a robust motion detection. This Paper deals with the fundamental concept of an entire system for classification of active motions and detection of passive postures for the application within a mobile ECG system which was developed at the Berlin Institute of Technology. The classification itself uses acceleration signals of a three-axis sensor located on the ECG chest strap. The motion detection is based on a generalized modeling. Therefore the system training does not need to be adapted for every new user individually. Another focus of this paper lies on the selection and validation of suitable representative features to achieve high specificity combined with adequate sensitivity for the generalized model. Finally effectiveness and efficiency of the designed system is investigated and the presented system is compared referring to other published systems for the detection of human motions.