Abstract - The control of powered upper limb prostheses using the surface electromyogram (EMG) is an important clinical option for amputees. There have been considerable recent improvements in prosthetic hands, but these currently lack a control scheme that can decode movement intent from the EMG to exploit their mechanical dexterity. Pattern recognition based control has the potential to decode many classes of movement intent, but is confounded when using the prosthesis in varying positions during activities of daily living. This work describes the degradation that can occur when using pattern recognition in varying positions, during both static positioning tasks and dynamic activities of daily living. It is shown that training with dynamic activities can greatly improve positional robustness for both static and dynamic tasks, without requiring a complex and lengthy training session. I. INTRODUCTION OWERED upper limb prostheses controlled using the surface electromyogram (EMG) have been available for many decades, allowing control of limb positioning and hand manipulation. The required muscular contractions are often similar to those needed to articulate an intact limb. Although it has been found that myoelectric prostheses can be clinically practical in upper limb prosthetics, , the limited dexterity of control is often cited as the primary reason for rather low acceptance of these devices [1]. Conventional myoelectric control schemes use an amplitude measure at each electrode site (such as the root mean square or mean absolute value of the EMG) to quantify the intensity of contraction in the underlying muscles. Control is elicited by mapping this activity to the desired prosthetic function. If more than one device is to be used, mode switching is often the only strategy, using a hardware switch or co-contraction to direct control to an elbow, wrist or hand. This method of control is often slow Manuscript received March 25, 2011. E. Scheme is a PhD candidate and Project Engineer at the Institute of Biomedical Engineering at the University of New Brunswick, Fredericton, NB Canada. (phone: 506-458-7029, fax: 506-453-4827) (email: escheme@unb.ca) K. Biron is a PhD candidate in the Department of Electrical Engineering at the University of New Brunswick, Fredericton, NB Canada.. (e-mail: katerina.biron@unb.ca). K. Englehart is Associate Director of the Institute of Biomedical Engineering at the University of New Brunswick, Fredericton, NB Canada. (e-mail: kengleha@unb.ca). This work was supported in part by NSERC Discovery Grant 217354, in part by NIH Grant 5R01HD058000-03, and in part by the Atlantic Innovation Fund. and counterintuitive. This has motivated the use of a pattern recognition approach to myoelectric control. By using multiple EMG sites, effective feature extraction and multidimensional classifiers, it is possible to discriminate many more classes of motion than with conventional control. The use of EMG pattern recognition has been shown to greatly improve the dexterity of control in upper limb prostheses and, through the efforts of many academic and commercial initiatives, it is nearing clinical viability [2]. There are a number of factors that currently challenge pattern recognition control in clinical settings, including variation in electrode placement [3,4] and impedance, and the effects of socket loading and limb position [5]. This work addresses the degradation caused by limb position during static and dynamic tasks, and describes an effective training strategy to minimize the effects. II. METHODOLOGY A. Experimental Protocol EMG data corresponding to eight classes of motion were collected from five right-handed, healthy, normally-limbed subjects (4 male, 1 female). All experiments were approved by the University of New Brunswick’s Research Ethics Board. The subjects were fitted with a cuff made of thermo formable gel (taken from a 6mm Alpha liner by Ohio Willow Wood) that was embedded with eight equally spaced pairs of stainless steel dome electrodes. The cuff was placed around the right forearm, proximal to the elbow, at the position with largest muscle bulk. A reference electrode was placed over the back of the hand. The eight channels of EMG were differentially amplified using remote AC electrode-amplifiers (BE328, by Liberating Technologies, Inc), and low pass filtered with a cutoff frequency of 500Hz. Finally, data were acquired using a 16 channel 16-bit analog-to-digital converter (USB1616FS from Measurement Computing TM ) sampling at 1kHz. Subjects were prompted to elicit a set of contractions consisting of the following eight classes of motion: wrist flexion/extension, wrist pronation/supination, hand open, power grip, pinch grip, and a no motion (i.e. rest) class. Subjects were encouraged to perform contractions at a repeatable ‘medium’ force level and given rest periods between trials to minimize fatigue. These sets were repeated during three sessions, each involving a different form of positional variation. Improving Myoelectric Pattern Recognition Positional Robustness Using Advanced Training Protocols E. Scheme, Student Member, IEEE, K. Biron, and K. Englehart, Senior Member, IEEE P