Conception and analysis of Cascaded Dual Kalman Filters as virtual sensors for mastication activity of stomatognathic craniomandibular system

The presented work shows a possible methodical approach for parameter estimation of a kinematic and dynamic element that characterizes a human mandibular system during the mastication process using position measurement only. The considered parameters are the velocity, friction coefficient, and the mass of the moving part of the mandibular during the mastication activity of a human. Internal or optical motion sensors can still allow imprecision in the measurements. To overcome these, in the present work a system identification algorithm is designed using a combination of three backward cascaded Kalman Filter, which consists of three Extended Kalman Filters. The identification procedure is validated through a matching criterion based on the estimation of the mass, which is assumed to be known in the first stage of the Kalman Filter structure. Three EKFs are tuned as long as the initial value of the mandibular mass is achieved as an estimation of the third one. This is due to the fact that the optimization procedure tries to optimize a non-convex optimization problem that can admit more than one solution. The main contribution of this project is designing state estimation dynamic system, which accurately estimates friction with a linear time varying model. Friction coefficient plays an important role in the early diagnosis of temporomandibular joints disorders, since it is very low under normal condition, and an increase may be associated with abnormalities. Computer simulations show the effectiveness of the proposed method to accurately estimate friction dynamics and refrain from complex nonlinearities.