Application of processing, which is specific to holographic concept with three reference signals, for complex reflection coefficient measurements by means of three-probe waveguide section is suggested. A mathematical model of diode detector on base of neural net is considered. The model provides a high level of square-law of output signal of detector and complex reflection coefficient measurements. The optimal model of neural net has structure of 1-4-1 neurons. For measurement of movement parameters estimation was done. As object of researches there was a metal concave and convex surface with various curvatures. For a decrease of a noise influence in the data, the Tikhonov’s regularization method was used. Efficiency of the method has been proved experimentally.