Modern radars must provide in a very short time: existence, mobility and shape of objects evolving in airspace. Evaluation of the object shapes through active research by using synthetic aperture radar is limited in time, resolution, and cost. A new way of processing non-stationary signals is presented in this article. Signals are obtained from the reflection of the electromagnetic field by objects with complex shape when they are irradiated with linear frequency modulated signals. The amplitude of reflected signal is variable on the radio-impulse duration depending on object shape, causing a certain electromagnetic signature. This phenomenon is caused by specific electromagnetic resonance. The reflected signal has maximum amplitude when the frequency of the incident wave is the same with the resonant frequency of the investigated object. The structure of an radar target can be decomposed into simple geometric shapes such as spheres, ellipsoids, prisms, and so on. Using resonant effect that ensures pattern recognition is exemplified by an object with an aerodynamic profile accepted in many component elements of the aircraft, namely - an ellipsoid. It is a geometric shape used extensively in aviation, because it has a very low aerodynamic resistance. The resonant response of ellipsoid is evaluated in a decade frequency band, but the pattern recognition of this shape is enough for an octave band. The resonant response is assessed for cross polarization of incident electromagnetic field, as well. As a result, the radio-impulse shape can be used in a data base for pattern recognition.