Detektion und Analyse physiologischer und biokinematischer Parameter mit Körpersensoren

The aim of this work is the acquisition of information about the health and physical performance of a human. Traditionally, clinical or laboratory environments with stationary measurement equipment are used for this purpose. However, technological advancements facilitate the development of compact, energy efficient microelectronic systems. Attached to the body of a human (body sensor), they allow information to be captured over long periods of time, in the common environment and of many people. Among others, possible applications are found within sport, occupational and preventive medicine as well as rehabilitation. This work contributes to the development of a body sensor. Covered are the extension, update and evaluation of the hardware platform, along with the realisation of computer programs for the analysis and visualisation of data. The body sensor is attached using a chest strap and the set-up does not require professional expertise. Therefore, this process is supported by the developed computer-based assistant, which monitors the signal quality. Subsequently, the focus lies on the development of algorithms to derive parameters from the sensor signals, which capture physical activities and the physiological responses. Established physiological parameters are the heart rate and its variability. With the developed hardware platform, the required detection of individual heart beats is possible with an accuracy of 97.9 %, even during high physical activity. The added sensor for measuring the thorax expansion and contraction allows the respiration rate to be determined with an average deviation from an ergospirometry system of -0.32 min-1. Furthermore, based on measurements of the acceleration sensor, a detection of walking and running velocities was developed. For the distinction of five velocities between 3 km-h and 11 km-h an accuracy of 97.8 % is achieved. Additional derived biokinematic parameters include the determination of the upper body tilt angles, the detection of steps in the acceleration measurements, the creation of activity profiles and the distinction of sport and daily activities. Moreover, it is shown how personspecific characteristics from the acquired biosignals can be extracted and combined to enable an identification of the person wearing the body sensor. The achieved accuracy ranges from 91.4 % to 98.1 %. Applications include the individualisation of an algorithm’s configuration as well as the interpretation of data by providing additional person-specific information. An example is the developed estimation of energy consumption, which requires the identity in addition to the derived physiological and biokinematic parameters. The parameters are therefore selected adaptively with regards to occurring signal disturbances. This work is further concerned with the optimisation of the energy requirement of the signal processing on microprocessors of body sensors. For this purpose, a software framework was created, which allows a design space exploration of the configuration parameters to be conducted. Determined are the Pareto-optimal configurations in terms of accuracy and energy requirement. Thereby, the energy consumption of, for example, the velocity detection can be reduced by the factor of 20, whilst the accuracy drops by only 1.7 %.