Files

Abstract

Hand amputation could dramatically degrade the life quality of amputees. Many amputees use prostheses to restore part of the hand functions. Myoelectric prosthesis provides the most dexterous control. However, they are facing high rejection rate. One of the reasons is the lack of sensory feedback. There is a need for providing sensory feedback for myoelectric prosthesis users. It can improve object manipulation abilities, enhance the perceptual embodiment of myoelectric prostheses and help reduce phantom limb pain. This PhD work focuses on building bi-directional prostheses for upper limb amputees. In the introduction chapter, first, an overview of upper limb amputee demographics and upper limb prosthesis is given. Then the human somatosensory system is briefly introduced. The next part reviews invasive and non-invasive sensory feedback methods reported in the literature. The rest of the chapter describes the motivation of the project and the thesis organization. The first step to build a bi-directional prostheses is to investigate natural and robust multifunctional prosthetic control. Most of the commerical prostheses apply non-pattern recognition based myoelectric control methods, which offers only limited functionalities. In this thesis work, pattern recognition based prosthetic control employing three commonly used and representative machine learning algorithms is investigated. Three datasets involving different levels of upper arm movements are used for testing the algorithm effectiveness. The influence of time-domain features, window and increment sizes, algorithms, and post-processing techniques are analyzed and discussed. The next three chapters address different aspects of providing sensory feedback. The first focus of sensory feedback process is the automatic phantom map detection. Many amputees have referred sensation from their missing hand on their residual limbs (phantom maps). This skin area can serve as a target for providing amputees with non-invasive tactile sensory feedback. One of the challenges of providing sensory feedback on the phantom map is to define the accurate boundary of each phantom digit because the phantom map distribution varies from person to person. Automatic phantom map detection methods based on four decomposition support vector machine algorithms and three sampling methods are proposed. The accuracy and training/ classification time of each algorithm using a dense stimulation array and two coarse stimulation arrays are presented and compared. The next focus of the thesis is to develop non-invasive tactile display. The design and psychophysical testing results of three types of non-invasive tactile feedback arrays are presented: two with vibrotactile modality and one with multi modality. For vibrotactile, two types of miniaturized vibrators: eccentric rotating masses (ERMs) and linear resonant actuators (LRAs) were first tested on healthy subjects and their effectiveness was compared. Then the ERMs are integrated into a vibrotactile glove to assess the feasibility of providing sensory feedback for unilateral upper limb amputees on the contralateral hand. For multimodal stimulation, miniature multimodal actuators integrating servomotors and vibrators were designed. The actuator can be used to deliver both high-frequency vibration and low-frequency pressures simultaneously. By utilizing two modalities at the same time, the actuator stimulates different types of mechanoreceptors and thus ha

Details

Actions

Preview