J Med Syst (2012) 36:841–851 DOI 10.1007/s10916-010-9548-2 ORIGINAL PAPER Identification of Hand and Finger Movements Using Multi Run ICA of Surface Electromyogram Ganesh R. Naik · Dinesh K. Kumar Received: 1 April 2010 / Accepted: 20 June 2010 / Published online: 7 July 2010 © Springer Science+Business Media, LLC 2010 Abstract Surface electromyogram (sEMG) based con- trol of prosthesis and computer assisted devices can provide the user with near natural control. Unfortu- nately there is no suitable technique to classify sEMG when the there are multiple active muscles such as during finger and wrist flexion due to cross-talk. Inde- pendent Component Analysis (ICA) to decompose the signal into individual muscle activity has been demon- strated to be useful. However, ICA is an iterative technique that has inherent randomness during initial- ization. The average improvement in classification of sEMG that was separated using ICA was very small, from 60% to 65%. To overcome this problem associ- ated with randomness of initialization, multi-run ICA (MICA) based sEMG classification system has been proposed and tested. MICA overcame the shortcoming and the results indicate that using MICA, the accuracy of identifying the finger and wrist actions using sEMG was 99%. Keywords Hand gesture sensing · Bio-signal analysis · Surface electromyography (sEMG) · Independent component analysis (ICA) · Bio-sensors · Blind source separation (BSS) · Multi run ICA G. R. Naik (B ) · D. K. Kumar School of Electrical and Computer Engineering, RMIT University, GPO Box 2476, Melbourne, VIC 3001, Australia e-mail: ganesh.naik@rmit.edu.au Introduction Hand and finger actions and maintained gestures are a result of simultaneous contraction of multiple mus- cles. There are numerous possible applications that are based on reliable identification of hand gestures includ- ing prosthesis control, human computer interface and games. There are three major modes of identification of the hand gestures; – mechanical sensors e.g.—sensor gloves [25] – vision data based on video analysis [27, 28] and – muscle electrical activity [6, 16] The use of glove requires an external device while the video sensing is dependent on lighting conditions and unsuitable for very small gestures. These two tech- niques are also not suitable for prosthesis control where the user may be an amputee and may not have a hand. Muscle electrical activity based system can be suitable for most trans-radial amputees and people with weak muscles as it does not require the actual movement but only the activation of the associated muscles. Surface Electromyography (sEMG) is the electrical recording of the muscle activity from the surface. It is closely related to the strength of muscle contraction and an obvious choice for control of the prosthesis and other similar applications. Many attempts have been made to use sEMG signal as the command to control the prosthesis [9, 30, 31], but to obtain a reliable command signal, the muscle needs to have high level of contraction and there should be only one active primary muscle. An example is the I-Limb [1] where despite the hand having individual finger control, the controller only has single muscle command. These techniques are not suitable for gestures where the muscle activity is