Movement intention detection for intuitive and non-intrusive prosthetic arm control

假肢有可能改善失去双臂的人的生活。然而，它们的功能是有限的，因为它们没有连接到人类神经系统。我的目标是通过检测人类的运动意图来设计和改进义肢和使用者之间的界面。当进行以目标为导向的运动时，人类使用眼睛和其他感官信息来指导手臂的运动。来自眼睛和手臂肌肉的综合信号将帮助我们检测运动意图。例如，在伸手抓取运动中，我们首先用眼睛扫描环境，识别目标，提取其位置、大小、形状和纹理，然后伸手抓取。然而，以往的运动意图检测算法大多是基于肌肉活动和大脑信号的。据我们所知，利用眼球测量来改善假体控制的研究很少。在眼球运动如何与运动意图相关，以及如何利用眼球测量和手臂肌肉信号的融合来改善假肢控制方面，这一知识空白需要填补。我的研究计划的长期目标是通过采用非侵入性和负担得起的技术来改进包括人体假体在内的辅助设备，并通过改进用户对运动意图的检测来更好地控制人造身体部分。我的短期目标(在接下来的5年内)是1)开发一种带有传感器的假肢手臂袜子来收集肌肉活动数据，2)检查肌肉和眼睛信号与运动意图之间的关系，3)使用包括眼睛跟踪技术在内的多模块输入来改进抓取检测，4)通过从多个生物信号通道进行机器学习并通过整合眼睛跟踪(瞳孔直径)来准确估计手指的运动，以及5)开发一个机器学习模型来准确预测连续的目标导向任务中的一系列运动意图。传感器融合技术也将在我的整个研究中实施，以利用多模块输入的优势。这项工作将产生将眼睛运动和肌肉活动信号转换为运动意图的重要工具，这是一种用于直观假肢控制的运动意图检测的创新方法。近年来，假肢装置有了很大的改进，但用于假肢控制的人机界面并没有充分发挥假肢的潜力。缩小这一差距将极大地促进假体系统的发展，并提高使用者的生活质量。我的研究将使大量失去肢体的人和假肢行业受益(增加就业和研究资金)。我期待从这项工作中产生的知识将使我们能够设计出智能假臂，并进一步提高我们对人机/机器人交互、远程操作和辅助设备控制的理解。

Prosthetic arms have the potential to improve the lives of individuals who have lost arms. Yet their functioning is limited as they are not connected to the human neural system. My goal is to design and improve the interface between a prosthetic arm and the user by detecting human movement intention. When performing a goal-directed movement, humans employ their eyes and other sensory information for guiding arm movement. Comprehensive signals from both eyes and arm muscles would help us to detect movement intention. For example, during a reach-and-grasp movement, we firstly use our eyes to scan the environment, identifying a target, extracting its location, size, shape and texture, and then reaching and grabbing. However, most previous algorithms for motion intention detection were based on muscle activities and brain signals. To the best of our knowledge, utilizing eye-metrics for improving prosthesis control has rarely been explored. This knowledge gap needs to be filled in terms of how eye-motions correlate to movement intention and how fusion of eye-metrics and arm muscles signals can be employed to improve prosthesis control. The long-term objective of my research program is to advance assistive devices including human prosthesis through the inclusion of non-invasive and affordable technologies, and by improving the detection of movement intention of the users for better control of the artificial body part. My short-term objectives (within the next 5 years) are to 1) develop a prosthetic arm sock with sensors to collect muscle activities data, 2) examine the relationship between muscle and eye signals to movement intention, 3) improve grasp detection using multiple module inputs including eye-tracking technology, 4) accurately estimate finger movement using machine learning from multiple channels of bio-signals and by integrating eye-tracking (pupil diameter), and 5) develop a machine learning model to accurately predict a series of movement intentions during a sequential goal-directed task. Sensor fusion technologies will also be implemented throughout my research to utilize the advantages of multiple modules input. This work will generate important tools to convert eye motion and muscle activity signals into movement intention, which is an innovative approach to movement intention detection for intuitive prosthesis control. Prosthetic devices have been greatly improved in recent years whereas man-machine interfaces for prosthesis control do not fulfill the device potential. Bridging this gap would greatly advance prosthesis systems and improve quality of life among users. My research will benefit both the large number of people who have lost limbs and the prosthesis industry (increase in jobs and research funding). I expect that the knowledge generated from this work will enable us to design a smart prosthetic arm and further improve our understanding of human-machine/robot interaction, remote manipulation and assistive device control.