An EMG-Driven Assistive Hand Exoskeleton for Spinal Cord Injury Patients: Maestro Youngmok Yun, Sarah Dancausse, Paria Esmatloo, Alfredo Serrato, Curtis A. Merring and Ashish D. Deshpande Abstract—In this paper, we present an electromyography (EMG)-driven assistive hand exoskeleton for spinal-cord-injury (SCI) patients. We developed an active assistive orthosis, called Maestro, which is light, comfortable, compliant, and capable of providing various hand poses. The EMG signal is obtained from a subject’s forearm, post-processed, and classified for operating Maestro. The performance of Maestro is evaluated by a standardized hand function test, called the Sollerman hand function test. The experimental results show that Maestro improved the hand function of SCI patients. I. I NTRODUCTION The number of spinal-cord-injury (SCI) patients is es- timated to be 282,000 in the United States in 2016 [1]. Approximately 45% of SCI patients have residual function in their arms and shoulders, but have difficulty performing activities of daily living (ADL) due to insufficient hand function. The goal of our research is to improve their hand function in ADL with an active assistive orthosis. Most current commercial assistive orthoses are passive devices that either help with passive extension/flexion or locate the fingers/thumb in a predetermined position [2]. Although these orthoses are economical and easy to use, they have several limitations. The passive stiffness or elasticity hinders hand movement when it is not needed. Moreover, they assume the subjects are able to apply enough force in at least one direction. In order to address these limitations, active orthoses have been recently developed. These devices recognize the intention of the subjects and assist them in achieving a task by adding extra strength. Since the active orthosis recognizes the intention, the assistive force is added only when a subject needs the force. In addition, because active orthoses add force, even a subject with weak muscles is able to perform tasks. There are two major challenges in developing an active assistive orthosis for hand function. First, the design of the device is challenging. The orthosis needs to be light and com- fortable to allow subjects to perform tasks while wearing it Y. Yun and P. Esmatloo are PhD students of Mechanical Engineer- ing, The University of Texas at Austin, USA yunyoungmok at utexas.edu S. Dancausse is a MS student of Mechanical Engineering, Ecole Nationale d’Ingénieurs de Saint-Etienne, France A. Serrato is a undergraduate student of Mechanical Engineering, The University of Texas at Austin, USA C. A. Merring is a OTR/L, MOT of Brain & Spine Recovery Center, Seton Brain & Spine Institute, USA A. D. Deshpande is a faculty of Mechanical Engineering, The University of Texas at Austin, USA Ashish at austin.utexas.edu for a considerable amount of time. The actuation of orthosis needs to be compliant. If the orthosis controls finger positions regardless of interaction force, it may harm the subject’s hand when interacting with a rigid object. The orthosis needs to provide various hand poses which are essential to perform various tasks in ADL. Benjuya and Kenney [3] pioneered the research of active hand orthosis. However, the geared- motor system on forearm resulted in a heavy system and non- compliant interaction. The orthosis introduced by DiCicco et al. [4] provided compliant interaction resulting from a pneumatic actuator, but it generated only pinching motion. The tendon-driven glove introduced by In et al. [5] is light and comfortable, but only capable of providing a wrap grasp. Recently developed exoskeletons [6], [7], whose primary purpose is rehabilitation, provide various hand poses, but most of these devices are not portable. The second major challenge is to recognize the intention of the subject. If an active orthosis fails to reliably identify the intention, it would actively hinder movements of subjects. One promising method is to use EMG signals for intention recognition. Since the EMG signals are obtained from task- relevant muscle groups, operation of the assistive orthosis is intuitive. Several researchers have developed active hand as- sistive orthoses driven by EMG signals for SCI subjects [3], [8]. However, the operations have been performed only for 1- DoF actuation and mainly with a binary threshold. Recently, Liu et al. [9] showed the potential of EMG signals of SCI subjects to predict their intended hand pose. They attached an EMG sensor array on the forearm of SCI subjects and built a map classifying the EMG signals into various essential hand poses of ADL. However, they have not implemented the classification algorithm for operation of any active assistive orthoses. In this paper, we present the novel design of an active assistive hand exoskeleton, called Maestro. Maestro is light, comfortable, compliant, and capable of providing various hand poses essential in ADL. Then, we present a reliable user-intention recognition method using EMG signals of SCI subjects. The method classifies EMG signals of hand muscles into intended hand motions, and sends a command to the controller of Maestro, leading to the desired hand pose of the SCI subject. Finally, the improvement of hand function in ADL is evaluated for two SCI subjects by a standardized hand function test, called Sollerman hand function test (SHFT) [10]. This paper presents a system level research including