A Compact Glove Input Device to Measure Human Hand, Wrist and Forearm Joint Positions for Teleoperation Applications M. A. Saliba, F. Farrugia, and A. Giordmaina Department of Manufacturing Engineering University of Malta Msida, Malta MSD 06 masali@eng.um.edu.mt Abstract—In this work, we have developed a new glove input device that is able to measure the angular joint positions of two fingers and of the thumb on the human hand, as well as the pitch position of the wrist and the roll position of the radio-ulnar joint of the human forearm. The glove has various new features, including the measurement of forearm roll position, that are not found in other glove input devices described in the literature. The glove contains a number of flexible, plastic bands whose displacement, during joint rotation, is measured using linear potentiometers. The new glove is light, compact, easy to wear and use, robust, and inexpensive, and is intended for use in teleoperation applications in conjunction with a remotely located robot hand/wrist. Another property is that it can be easily adjusted to fit a wide range of human hand sizes. Preliminary testing of the glove has shown that it can achieve an accuracy in position measurement that compares well to that of a number of commercially-produced gloves that are presently in use. Keywords—glove input device; whole hand input device; teleoperation; human joint position sensing. I. INTRODUCTION The development of data gloves, or whole hand input devices, for the sensing of human hand and wrist movements, is an area that has received a fair amount of attention over the last twenty five years. One of the major commercial exponents of this line of research and technology has been the entertainment industry, specifically in the area of Virtual Reality (VR), where the use of a whole hand input device allows the user to experience a more transparent interaction with a computer- generated environment. However, data gloves also have a number of other very compelling applications outside of VR. These include, for example, the teleoperation of multi degree-of-freedom anthropomorphic robot hands for use in remote and/or hazardous environments; the programming of complex human-type hand motion sequences for autonomous, dexterous robot hand applications, such as for material handling in industry; the implementation of “point, reach, and grab” commands in a gestural interface; and the interpretation of human hand gestures for use by the vocally impaired. The very earliest data gloves were designed to replace the keyboard as a text input device. These involved fitting the human hand into or against a glove-like device, such that the thumb and each finger could be used to activate a unique switch, and the five signals were used to encode the various character symbols [1] [2]. The gloves in [3] were fitted with electrical contacts on the fingertips, and were designed to work in conjunction with a modified keyboard, fitted with exposed electrical contacts on the upper surface of the keys, as an educational device for teaching touch typing skills. Possibly the earliest whole hand input device with generic applications was the digital data entry glove in [4]. This glove was intended both as a substitute for the keyboard and mouse for character and command entry into a standard computer, as well as for the interpretation of hand positions representing the Single Hand Manual Alphabet used by the vocally/aurally impaired. The glove comprised a number of touch/proximity sensors (to detect contact between the fingers, or between the fingers and the palm), a number of “knuckle-bend” sensors (to detect flexing of specific hand joints), a number of tilt sensors (to detect hand tilt through the horizontal plane), and a number of inertial sensors (to detect hand acceleration in two orthogonal directions). Each of the sensors gave a two-state output, so that the interpretation of a hand gesture involved the logical decoding of the specific combinations of the digital output signals. The first commercial glove input device was the VPL DataGlove, developed by VPL Research Inc. [5]. This glove utilizes various sensor technologies, including ten fibre-optic cables that pass over the first two joints of each of the five digits of the human hand, and that exhibit reduced light transmission when they bend, as a function of joint angle [6]. Evaluations of the performance of the VPL DataGlove show the mean error in the sensed position of the fingers to be about 6 degrees (that of the thumb is substantially higher) [7] [8]. A more recent commercial glove input device is the CyberGlove by Immersion Corporation, which uses up to twenty two resistive bend sensors to measure the angular position of the joints of the hand [9]. An evaluation of the glove [10] showed sensor accuracy to be considerably better than that of the VPL DataGlove, with errors of only a few degrees for most of the joints, and with the thumb joints once again displaying the greatest errors.