A basic study of feedback parameters from AR model for the biofeedback system aiming human tremor control

Abstract

The purpose of this study is to clarify what kind of bio-information must be fed back, when Parkinson's disease and essential tremor patients use biofeedback training for their tremor control. We applied the AR (autoregressive) model to the forearm tremor in order to have a clearer picture of their features from the viewpoint that information related to the difference between pathological tremors and physiological tremors of healthy people may be fed back. Subjects are 18 Parkinson's disease patients, 20 essential tremor patients, and 13 healthy old people. We used the 7-th degree AR model which minimizes FPE (final prediction error'). The following resuls were obtained : (1) the characteristic equation obtained from the pathological tremor observed in Parkinson's disease and essential tremor patients contains mostly negative real root as a factor (on the other hand, the one obtained from the physiological tremor of healthy people contains mostly positive real root); (2) within the roots of characteristic equation obtained from the pathological tremor of essential tremor patients, the absolute value of characteristic root which shows fundamental frequency is mostly closer to 1 (on the other hand, within the roots of characteristic equation obtained from the physiological tremor of healthy people, the one is mostly far from 1). We will train patients with the tremor diseases, while those differences obtained here are presented to them as biofeedback information. We can use the result obtained in this study as it is for the diagnosis support of the Parkinson's disease and the essential tremor.