Estimation of multiple frequencies in pathological hand tremor for a semi-active 2DOF dynamic vibration absorber designed for tremor suppression

Abstract

Pathological tremor is the most prevalent movement disorder in the world, where ~14.5% of people aged 50-89 years suffer from it. The most common types of tremor are by far essential tremors (ET) and Parkinson Disease (PD). Many people experience the shaking as embarrassing, and Social Anxiety Disorder is a common occurrence among people with ET. Today's treatments for tremors consists mainly of medication or neurosurgery, but unfortunately, tremor is not adequately managed in ~25% off all patients.

Our end goal is to reduce tremors in the upper limb with a mechanical solution in a non-embarrassing way. In cooperation with Ida Estenstad a design for a vibration absorber for reducing pathological hand tremor was developed. The system is semi-active, and the design consists of a dual parallel vibration absorber that can tune into two frequencies in the hand tremor. For the system to be able to work in real time, an estimator that can track these frequencies is needed. This thesis presents the development and validation of a frequency estimator for tracking multiple frequencies in pathological hand tremor for real-time applications; the Band-limited Multiple Weighted Fourier Linear Combiner (BMWFLC). Multiple tests with real and simulated tremor data was carried out to validate its performance. Using a simulated signal of a rest tremor modulated by two frequencies, the filter had two stable estimates in under 3 seconds with an error under 0.053 Hz for both. The filter showed good performance when tested on real data from both action and rest tremor. The algorithm was also implemented on an 8-bit microcontroller and tested on a 1DOF test rig that simulates pathological hand tremor with flexion-extension movement. The algorithm ran in real-time, but not with the ideal resolution or frequency window size, and only tracked one frequency. An upgrade of the microcontroller is suggested to increase the performance.

To the authors' knowledge, this is the only frequency estimator for pathological tremor estimation that can estimate frequencies that are close in the frequency domain and the first with the capability to estimate over two frequencies simultaneously. \newline