Algorithmusunterstützte Multisensorintegration zur langzeitstabilen Objektverfolgung und Bewegungserkennung

Since the required hardware for low-cost inertial navigation is very small, low-weight and widely available in the market, using MEMS-based sensors promises various industrial, medical or consumer entertainment applications to be realized at very low manufacturing costs. Due to the performance these sensors have shown, their applicability such as shaking detection, vibration measurement or orientation-dependent screen view rotation, has been limited to just simple, non-intelligent tasks in smart phones, tablets etc.
The present research is motivated by enhancing the currently available applicability of low-cost inertial navigation systems (also: INS), towards intelligent applications of long-term object tracking and motion recognition.
The object tracking makes use of a loose coupling integration method based on Kalman-filtering in order to realize a sensor fusion between low-cost INS and GPS reference navigation. This work shows the performance of two-filter-smoothing to reduce the growth of errors during potential outages of the reference navigation. A simplification technique is applied to avoid the calculation of inverse covariance matrices for the smoothing, which reduces the possibility of numerical instabilities while increasing the algorithms’ efficiency.
Aiming towards a reliable and simple possibility of computer-assisted motion analysis and validation as a key for motion optimization, the present work provides an algo-rithmic framework for reference-less motion analysis and validation using low-cost inertial sensors. The developed algorithms are based on the theory of Hidden-Markov-Models and on stochastical modelling of inertially measurable motion profiles using Markov-chains.
The research results are validated by a series of experiments in order to verify the reliability and the stability of the present approaches. The final solutions are stand-alone, low-cost, miniature-size and low-weight, while being capable of unlimited long-term operation.