This thesis investigates methods of improving myoelectric prosthetic hand control as a complete human-in-the-loop control system via a multifaceted approach; in terms of the range of contributions to core myoelectric control elements, and also the diverse examined physical and psychological effects. Myoelectric prostheses have been the subject of over 75 years of attention in research, development, and even fiction. Despite this, user rejection is common, with issues such as discomfort and functionality as prevalent now as they were decades ago. This thesis aims to address these problems by challenging established ideologies and by examining key aspects of the entire human-in-the-loop system, equally weighting user training, end-effector design, and control algorithm development, and evaluating the physical and psychological effects to a more complete level than is seen in existing literature.

First, myoelectric prosthetic hand control is introduced and literature reviewed. Next, a robot-enhanced virtual reality platform for prosthetic hand training is presented, which demonstrates higher quality and more representative training than virtual reality alone. Chapter three outlines two personalised discrete myoelectric controllers with comparative performance to computationally heavy controllers, the second achieving near-instant calibration from minimal data. Chapter four presents a closed-loop continuous myoelectric controller, which combines a proprioceptive feedback armband with a multi-degree-of-freedom continuous controller which requires minimal user data, leading to improvements in isolated control and sensory tasks, precision grasping, and faster psychological embodiment. Chapter five examines prosthetic hand design itself, and presents grasp-specific and task-specific prostheses; the former achieves power and precision grasps with minimal complexity, while the latter dramatically outperform a conventional prosthesis on targeted tasks.

The goals of this thesis are to encourage practitioners to view myoelectric prosthetic hand control as the complex, human-centred system that it is, and to provide key starting points toward improving myoelectric control, and ultimately improving end-user outcomes.