NONLINEAR INTERPOLATION OF MANDIBULAR KINESIOGRAPHIC SIGNALS BY APPLYING SENSITIVITY METHOD TO A GMDH CORRECTION MODEL

The purpose of the present paper was to propose an improved nonlinear interpolation method for estimating distorted kinesiographic recording of interlattice points in space, and to evaluate its correction accuracy. The group method of data handling (GMDH) correction method we have reported previously is essentially different from other geometric correction methods in the sense that the GMDH correction model can be adapted to the measurement environment in which kinesiographic signals are subject to ferromagnetic interferences. Due to its inherent structure, however, our previous GMDH correction model was restricted to kinesiographic signals representing lattice points in space. In this paper, we explored the sensitivity method for GMDH correction modeling. This provides for nonlinear interpolation of kinesiographic signals for interlattice points. Nominals, i.e., true values, of 3-D coordinates of the interlattice points were determined by means of a mandibular jaw movement simulator together with simultaneous recording of distorted kinesiographic signals which corresponded to the nominals. Distorted signals were corrected by the new method of correction modeling. A mean estimation error of 0.16 mm (SD 0.19 mm) was determined for 24 interlattice coordinates. Thus, nonlinear interpolation by the sensitivity method was confirmed to be effective.