Intent Seeking Algorithms for New Human-Machine Interface

项目摘要：我们为瘫痪患者创造了一种新的人机界面（HMI）技术，该技术利用单个面部肌肉的表面肌电图（sEMG）信号同时多维地控制外部设备。我们的新控制器有潜力显著提高用户的生活质量。与许多现有的针对这一人群的人机界面不同，我们的界面是：不显眼，不干扰眼睛/嘴/舌头，在需要时持续可用，多功能，几乎在任何头部位置都易于使用，并且便携。我们最近发现，人类可以学会如何同时操纵表面肌电信号功率谱的两个不同频段的功率水平（只需通过收缩肌肉）。每个频段成为一个单独的控制通道，可以同时控制设备的不同方面。因此，我们可以利用单个肌肉？以比以前所知的复杂得多的方式研究自然电信号。利用这一潜在的发现，我们开发了一种新的用户界面，它依赖于单个头部肌肉？表面肌电信号，这是很容易获得和限制在一个小的不可描述的区域靠近耳朵。我们的系统在某种程度上类似于一些基于脑电图（EEG）的脑机接口（BCI），在这种接口中，一个人学习如何引导？在计算机屏幕上指向特定位置的光标。屏幕上的这些位置可以是打开计算机应用程序（人机界面）、开/关灯（环境控制单元）或控制轮椅（移动应用程序）的虚拟按钮。两个主要的挑战：光标引导？HMI系统是1)意图（计算机如何知道用户打算把光标放在哪里？）和2)光标到达预期位置的速度。这两个问题是交织在一起的，因为早期的意图知识可以导致更快的系统。我们建议发展新的“意向导向”。这些算法可以使我们的人机界面比我们目前实例化的系统快得多。此外，为了对不能离开家或医院的残疾人进行评估研究，我们将开发一种新的更小的移动版硬件，它非常易于运输、安装和在任何地方使用。智力优势：在人机界面中使用单个表面肌电信号同时进行多维控制具有潜在的变革性。预测目标（在本例中是计算机光标）未来位置的概念出现在许多不同的应用中（例如航空航天工程、机器人技术、脑机接口）。这些应用结合了数学和计算机科学技术，包括统计决策、最优过滤和人工智能。我们打算借鉴这些领域，为我们的人机界面应用开发准确、快速的算法。从硬件的角度来看，全新的计算设备正在出现，它们可以在手持（或更小的）设备上执行复杂的计算并运行图形密集型应用程序。围绕这些操作平台设计我们的界面将推进高度便携和易于使用的辅助界面领域。更广泛的影响：最近由里夫基金会发起的一项研究（2009年）估计，在美国有超过550万人患有瘫痪。许多最严重的瘫痪患者使用呼吸机呼吸，并且在一天中的不同时间被限制在特定的头部/身体位置。我们的目标是让严重瘫痪的人重新获得对周围环境的控制和一些基本的独立性。我们致力于将残疾人纳入我们的研究，不仅作为研究对象，而且作为研究人员本身。因此，我们的工作将在研究包容性方面产生更广泛的影响。就更广泛的智力影响而言，我们的新意图搜索算法可以应用于许多计算机操作系统/程序，残疾人或非残疾人使用各种设备来引导屏幕上的光标。

PI: Joshi, Sanjay S.Proposal Number: 0966963Project Summary: We have created a novel human-machine interface (HMI) technology for paralyzed persons which uses the surface electromyography (sEMG) signal of a single, facial muscle for simultaneous multidimensional control of external devices. Our new controller has the potential to significantly increase the quality of life for its users. Unlike many existing humancomputer interfaces for this population, our interface is: unobtrusive & inconspicuous, noninterfering with eyes/mouth/tongue, continuously available when needed, multifunctional, easy-to-use in almost any head position, and portable. We have recently discovered that humans can learn how to simultaneously manipulate power levels in two separate frequency-bands of a sEMG power spectrum (simply by contracting the muscle). Each frequency band becomes a separate control channel, which can simultaneously control different aspects of a device. Thus, we may exploit a single muscle?s natural electrical signals in far more complex ways than previously known. Using this underlying discovery, we have developed a new user interface that relies on a single head muscle?s surface EMG signal, which is easy to obtain and restricted to a small non-descript area near the ear. Our system is somewhat similar to some electroencephalographic (EEG) based brain-computer interfaces (BCI) in which a person learns to ?guide? a cursor to certain positions on a computer screen. These positions on the screen could be virtual buttons that open computer applications (human-computer interfaces), turn on/off lights (environmental control units), or control wheelchairs (mobility applications). Two central challenges in all ?cursor-guided? HMI systems are 1) intent (how does the computer know where the user intended to place the cursor?), and 2) speed at which the cursor can achieve the intended position. These two questions are intertwined in that earlier knowledge of intent can lead to faster systems. We propose to develop new ?intent-seeking? algorithms that could make our HMI much faster than our currently instantiated system. In addition, in order to conduct evaluation studies on subjects with disabilities who cannot leave either home or hospital, we will develop a new smaller mobile version of our hardware that is very easy to transport, setup, and use anywhere.Intellectual Merit: The use of a single sEMG signal for simultaneous multidimensional control in human-computer interfaces is potentially transformative. The notion of predicting the future location of a target (in this case a computer cursor) arises in many different applications (e.g. aerospace engineering, robotics, brain-computer interfaces). These applications employ a combination of mathematical and computer-science techniques including statistical decision making, optimal filtering, and artificial intelligence. We intend to draw from these fields to develop accurate, fast algorithms for our human-computer interface application. From a hardware perspective, entire new classes of computing devices are appearing that can perform complex computations and run graphics-intensive applications from a hand-held (or smaller) footprint.Designing our interface around these operating platforms will advance the area of highly portable and easy-to-use assistive interfaces.Broader Impact: A recent study initiated by the Reeve Foundation (2009) estimates that more than 5.5 million people live with paralysis in the United States. Many of the most severely paralyzed use ventilators to breathe, and are confined to certain head/body positions at different times during the day. Our goal is for severely paralyzed persons to regain some control of their surroundings and some basic independence. We are committed to including disabled persons in our research, not only as subjects but also as researchers themselves. As such, our work will create an additional broader impact in terms of research inclusiveness. In terms of intellectual broader impact, our new intent-seeking algorithms could have applications for many computer operating systems/programs for which disabled or non-disabled persons use various devices to guide cursors on a screen.