Multisensor fusion with a pattern recognition approach: parametric case

This paper describes a fusion algorithm based on a pattern recognition approach. It is concerned with the case of a time-invariant parameter that may have multiple values when the a priori knowledge about the number of values is incomplete. With a pattern recognition approach, the sensor fusion problem is a twofold problem: first to recognize the values of the parameter and to classify the measurements with a discrimination rule and second to estimate these values taking measurements accuracy into account. Finally, the result of the sensor fusion is identified to a multimodal probability density function.