GMDH Correction Modeling of Distorted Signals Recorded by Mandibular Kinesiograph
Motoyasu Nagata, Kenji Takada, Mamoru Sakuda · IEEE Transactions on Biomedical Engineering · 1986
A new correction method is investigated for distorted measurement signals recorded by a mandibular kinesiograph, an electromagnetic device to quantify jaw movement by a magnetic transducer fixed on the mandible. The method employs a self-organized system identification known as a group method of data handling (GMDH). The proposed GMDH correction method has the capability to design a two-dimensional nonlinear autoregressive model which represents the relationship between distorted measurements and their corrected estimates of the mandibular displacement for the initial adjustment of kinesiographic signals under ferromagnetic influences. A computational procedure of a GMDH correction method is described in which the mathematical structure of a nonlinear autoregressive model is identified in a self-organized manner. Heuristic determination of a restricted sensor array zone of a kinesiograph is proposed to obtain effective fitness of a particular GMDH correction model which corresponds to each of the concomitant restricted zones. Corrected signals evaluated in the present study reveal a mean estimation error of 0.101 mm with a standard deviation of 0.125 mm which is regarded as highly accurate in a clinical sense.