2168-2194 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JBHI.2014.2340397, IEEE Journal of Biomedical and Health Informatics JBHI-00255-2014R1 1 Abstract— Identification of simple and complex finger flexion movements using Surface Electromyography (sEMG) and muscle activation strategy is necessary to control Human Computer Interfaces (HCI) such as prosthesis and orthoses. In order to identify these movements, sEMG sensors are placed on both anterior and posterior muscle compartments of the forearm. In general, the accuracy of myoelectric classification depends on several factors, which include number of sensors, features extraction methods and classification algorithms. Myoelectric classification using a minimum number of sensors and optimal electrode configuration is always a challenging task. Sometimes, using several sensors including high density electrodes will not guarantee high classification accuracy. In this research we investigated the dependency and independency nature of anterior and posterior muscles during simple and complex finger flexion movements. The outcome of this research shows that posterior parts of the hand muscles are dependent and hence responsible for most of simple finger flexion. On the other hand this study shows that anterior muscles are responsible for most complex finger flexion. This also indicates that simple finger flexion can be identified using sEMG sensors connected only on anterior muscles (making posterior placement either independent or redundant), and vice versa is true for complex actions which can be easily identified using sEMG sensors on posterior muscles. The result of this study is beneficial for optimal electrode configuration and design of prosthetics and other related devices using a minimum number of sensors. Index Terms— Surface Electromyography (sEMG); Subband Decomposition ICA (SDICA); Blind Source Separation (BSS); Anterior; Posterior; simple and complex flexion. I. INTRODUCTION urface Electromyography (sEMG) represents the level of muscle activity recorded from the skin surface. It provides rich motor control information and is closely related to the Manuscript received April, 11, 2014,―This work was supported in part by the UTS Chancellor‘s Postdoctoral Fellowship Grant‖. Ganesh R. Naik and Hung T Nguyen are with Centre for Health Technologies (CHT), University of Technology Sydney, 2007, Australia (e- mail: Ganesh.Naik@uts.edu.au; Hung.Nguyen@uts.edu.au). Kerry G. Baker is with School of Medical and Molecular Biosciences Faculty of Science, University of Technology Sydney, 2007, Australia (Kerry.baker@uts.edu.au). strength of muscle contraction [1, 2]. In the recent past myoelectric signals were extensively used for prosthetics [3- 6], wheel chairs [7, 8], exoskeleton robotics [9, 10], silent speech recognition [11] and rehabilitation applications [9, 12]. Myoelectric classification depends on several factors which include electrode selection [13], placement of electrodes [14- 17], feature extraction methods, selection of appropriate classifier algorithms [18] and computational complexity associated with myoelectric classification [19, 20]. Researchers have been working extensively to improve the myoelectric classification accuracy by improving the above factors; however, cross talk and noise makes it difficult to achieve higher rate of recognition. The most significant elements that contribute to the amount of detected crosstalk signal are: (i) sensor placement on the surface of the muscle and (ii) the spacing between the electrodes on the sEMG sensor [21]. The electrode placements and effect of electrode shift on sEMG pattern recognition has been previously investigated with varied results [22, 23]. In a recent study, Hargrove et al. [23], found different results using five electrodes that are connected parallel to the muscle fibers. Another study conducted by the same authors investigated the placement of electrode poles and concluded that transverse orientation of electrodes are more sensitive to shift than longitudinal orientation [14]. A previous version of our proposed method on sEMG electrode sensor placement concentrated mainly on simple gestures [15]. However, issues remain to be resolved such as selection and placement of electrodes for identification of simple and complex gestures [1, 18]. Anterior and posterior hand muscles are responsible for simple and complex finger flexions and actions. However, it is known phenomenon that while muscles in the anterior compartment are contracting there is co-activation of muscles in the posterior compartment [24, 25]. In general, these (anterior and posterior) muscles are not contributing to flexion but are impeding movement to better control the action by muscles in the anterior compartment [24]. Hence, there is a need for proper signal processing and pattern recognition methods, which can evaluate and identify approximate location for placement of electrodes in identifying gestures. By doing the above, we can identify different simple and Dependency Independency measure for posterior and anterior EMG sensors used in simple and complex finger flexion movements: Evaluation using SDICA Ganesh R. Naik, Member, IEEE, Kerry G. Baker, and Hung T. Nguyen, Senior Member, IEEE S