Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Engineering Science

Program

Electrical and Computer Engineering

Supervisor

Lyndon J. Brown

Abstract

Parkinsonian tremor is one of the clinical hallmarks of Parkinson's disease. Since the traditional medical treatments are not effective, many wearable devices are developed to help suppress the tremor. In order to suppress the tremor, a well-designed tremor estimator is needed. Previous tremor estimators treat a 3-D tremor signal as three independent 1-D signals. Moreover, they did not consider the real-life characteristics of tremor signals. For instance, the tremor does not always exist in the postural tremor signal, and the patient's voluntary motion can be included in the kinetic tremor signal. This paper presents a real-time adaptive parkinsonian tremor signal identifier based on the internal model principle and instantaneous Fourier decomposition and tests on tremor signals collected by a special glove from 18 patients. The result showed that our proposed identifier could identify a 3-D tremor signal and have the ability to recognize the presence of tremor and separate the voluntary motion from the tremor signal. We also showed that our proposed identifier could achieve 80%+ in signal identification accuracy and 90%+ in power estimation accuracy in different tremor signals. Finally, we achieved real-time tremor identification in a bench-top tremor simulator.

Summary for Lay Audience

Parkinson's disease (PD), also known as shaking palsy, is a neurodegenerative disease commonly developed in people older than 60. PD's signs and symptoms may include Parkinsonian tremor, slowness of movement and rigidity. Unfortunately, there are no medical therapies that can completely cure PD nowadays. Therefore, researchers are already looking to improve symptoms through external treatments like wearable devices.

Among all symptoms, the Parkinsonian tremor, or shaking, is the most prominent and troublesome symptom in patients' daily life. In recent years, many wearable devices are being invented to help suppress the tremor. In order to achieve good tremor suppression, it is necessary to achieve good tremor identification priorly. This research presents a Parkinsonian tremor signal identifier to achieve real-time tremor identification on the 3-D signals collected by a specially designed glove from 18 patients' hands. The simulation results showed that our proposed identifier could achieve expected accuracy in resting tremor signals, which were recorded when patients' hands were completely resting against gravity.

However, in real-life scenarios, the tremor may exhibit different characteristics. Moreover, it may become less noticeable or disappear entirely when patients try to control their hands to do voluntary movements. Hence, we improved the identifier and granted its ability to identify different types of tremors, recognize the tremors' presence and disappearance and separate voluntary motion from tremor signal. The simulation results showed that our proposed identifier could maintain high identification performance in postural and kinetic tremor signals, which were recorded when patients were asking to do different tasks with their hands. Finally, we achieved real-time tremor identification in an artificial tremor simulator.

Download

Share

COinS