Abstract—This paper presents the design of an assistive real-time system for the upper limb disabled to access a computer via residual muscle activities without standard computer interfaces (e.g. a mouse and a keyboard). For this purpose, electromyogram (EMG) signals from muscles in the lower arm were extracted and filtered using signal statistics (mean and variance). In order to control movement and clicking of a cursor from the obtained signals, six patterns were classified, applying a supervised multi-layer neural network trained by a backpropagation algorithm. In addition, an on-screen keyboard was developed, making it possible to enter Roman and Korean letters on the computer. Using this computer interface, the user can browse the Internet and read/send e-mail. The developed computer interface provides an alternative means for individuals with motor disabilities to access computers. A possible extension of our interface methodology can be incorporated in controlling bionic robot systems for the limb disabled (e.g. exoskeletons, limb prostheses). I. INTRODUCTION o date, many researchers have developed alternative interfaces to allow the upper limb disabled to access computers. Recently, neural signals have been attracting attention with respect to extracting user’s intention, as these signals provide information related to body motion faster than other means (e.g. kinematic and dynamic interfaces). Notably, a variety of methods have been developed to execute a user’s intention: from brain or muscle activities. At the central nervous system (CNS) level, signals from brain activities are applicable candidates to extract human thoughts. The electroencephalogram (EEG) [1] is a non-invasive monitoring method to record brain activities on the scalp. However, the signals acquired via this method are massed activities of many cortical neurons, and provide low spatial resolution and a low signal to noise ratio (SNR). On the other hand, invasive monitoring methods capture the activities of individual cortical neurons in the brain [2]. Manuscript received February 8, 2007. This work was supported by grants from Korea Institute of Science and Technology (KIST) and Brain Korea 21 (BK 21) Project of the Ministry of Education and Human Resources Development. Changmok Choi is with School of Mechanical, Aerospace & Systems Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea (corresponding author to provide phone: +82-42-869-3271; fax: +82-42-869-5230; e-mail: axlguitar@kaist.ac.kr ). Jung Kim is an assistive professor with School of Mechanical, Aerospace & Systems Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea (corresponding author to provide phone: +82-42-869-3231; fax: +82-42-869-5230; e-mail: jungkim@kaist.ac.kr ). However, many fundamental neurobiological questions and technical difficulties need to be solved [3], and interface methods based on brain activities generally require extensive training [4]. Despite these challenges, research in this area shows promise for helping people with severe motor disabilities (such as loss of skeletal muscle control from below the shoulders). A standard signal in a peripheral nervous system (PNS) level is the electromyogram, (EMG) [5] which represents muscle activations. EMG signals can be measured more conveniently and safely than neural signals at a CNS level. Furthermore, this non-invasive monitoring method provides good SNR. Hence, EMG-based HCI implementation is more practical with current technology. This paper presents an EMG-based computer interface that enables the upper limb disabled, such as quadriplegic (C7, C8 functional level) patients and hand amputees, to access a computer without standard computer interfacing devices (e.g. a mouse and a keyboard), as depicted in Fig. 1. Using the developed computer interface, users can alternatively control movement of cursor and click buttons through muscle activation in the lower arm. Also, using the designed on-screen keyboard, they can enter Roman and Korean letters on the computer. In order to confirm the utility of the developed computer interface, an experimental study was conducted to evaluate performance by applying Fitts’ law, which is a model that quantitatively evaluates the effectiveness of a computer pointing device. While some researchers have presented similar computer interfaces from EMG signals [6, 7], they have focused mainly on implementation of the interface and have not quantitatively A Real-time EMG-based Assistive Computer Interface for the Upper Limb Disabled Changmok Choi and Jung Kim T Fig. 1. Conceptual diagram of the developed EMG-based computer interface 1-4244-1320-6/07/$25.00 (c)2007 IEEE 459 Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands