Hervé Nguendon Kenhagho

Werdegang

sources, mainly ultra-wideband reflection data. This includes filtering procedures for the attenuation of white noise components with different filter types as IIR and FIR. In order to preserve distinctive features, the usage of Savitzky-Golay filter is an appropriate approach to follow. In addition, dependencies between single features remain to be detected by methods of multivariate regression analysis. Classification and segmentation of data sets needs to be examined by techniques as k-means clustering or SVM. At the end of the chain of preprocessing, single emission data or emission vectors relating to certain states are identified. This information is used to estimate the most likely state or resp. situation. This may be an activity of the user or a state of health. Most of the situations are described and represented by stochastic automaton state or a sequence of single automaton states. The stochastic behavior of these state sequences can be naturally expressed by discrete
distributions or continuous distributions. This included also continuous or discrete distributions for the symbols emitted in each state. The detection of states is indispensable for the interpretation of the sensor raw data of the ultra wide band radar. The fusion of raw data resp. radar reflection and context information is the precondition for a reliable, unobtrusive acquisition of user related data – the main aim in research. The implementation and composition of techniques for raw data analysis, information processing, sensor fusion, situation recognition and context aware radar raw data interpretation for feature extraction is the main aspect of research. A current approch od data fusion is based on  the Semantic Sensor Network by using ontology standard process engineering.