Response to fast speed stimuli of non linear system is difficult to predict without a proper model whose parameter has been characterized in condition closer to the working ones, because assumptions and simplifications on component stiffness and damping, such as their linearity in the frequency domain, are no longer valid. To characterize a two degrees of freedom spring-mass-damper system reacting to a high speed impact a set of statistical techniques has been developed and applied to a set of experimental measurements obtained thanks to a high speed camera, a piezoelectric load cell and a set of piezoaccelerometers. Photogrammetric position accuracy has been improved using a least square estimator of a single position from different markers, therefore position and acceleration data have been fused together being weighted on the reciprocal of their uncertainties. Starting from load, position, velocity and acceleration data, spring-mass-damper systems with different degrees of freedom have been characterized in different configurations using genetic algorithms to minimize the square deviation of prediction from measured speed, although the term to be minimized has been again weighted on reverse of uncertainty to avoid false results. Eventually the whole process has been validated and automated in order to provide a new and efficient mean of characterizing dynamic parameters such as stiffness, damping, trigger length of elastic systems impacted by high speed 1 objects.