In this project we will explore the use of a novel prosthesis controller based on the principle of Principal Component Analysis to enable seamless posture selection in high degree-of-freedom (DOF) prosthetic hands. The goal of this project is to develop a multi-degree of freedom (DOF) hand prosthesis posture controller that uses myoelectric signals (EMG) as control inputs and which has been dimensionally optimized using principal component analysis (PCA). Currently available multi-DOF hand prostheses cannot be fully utilized because there are fewer control inputs than the number of DOFs to be controlled (i.e. an underactuated system). Based on work from the neuroscience literature1 it has been shown that grasping is a 'low dimensional' task. This work used PCA to quantify the principal components (number of dimensions) involved in grasping. It was found that grasping tasks could be well described by the first two principal components. Two principal components implies that the posture of a multi-DOF hand, while grasping, can be controlled using only 2 degrees-of-control. This is an encouraging finding since current clinical upper limb prosthetic practice indicates only 3 or 4 independent myoelectric sites can be located on the residual limb of a typical person with a transradial amputation. We propose to explore the merits of a hand posture controller based on the first two principal components described by Santello et al.1 and driven using 2, 3 or 4 myoelectric sites. Santello et al. measured 15 joint angles in the hand of the subjects while 'grasping' 57 household objects. The resulting analysis showed a high amount of covariance between the joints while grasping different objects. A principal component analysis showed that the first two principal components accounted for 84% of the variance. This result suggests that, for grasping tasks, control of our 22 DOF natural hand reduces to a largely 2 dimensional control problem. Applying this finding to the control of multi-articulated prosthetic hands means we can use 2-4 myoelectrodes yet still be able to seamlessly move between postures in a multi-DOF hand. We will develop a control algorithm that will map the myoelectric signals to weighted combinations of Santello et al.s first two principal components to yield a desired posture. All functional grasp as defined by Keller et al., (1947) are achievable by varying the degree to which either principal component is weighted. This controller will direct high-dimension grasps with only 3 or 4 myoelectric sites and therefore control a multi-degree of freedom prosthetic hand using currently available clinical practices. This is of relevance because there are a number of new commercially available hands coming onto the market with articulated fingers and multi- positional thumbs - but with no way to select between grasps in a easy manner. We will demonstrate an EMG- driven PCA-based controller by having it drive a Bebionic Hand2 which has been modified to have a two degree-of-freedom thumb - converting it into a 6 DOF hand. 1 Santello, M., Flanders, M., and Soechting J.F., (1998): Postural hand synergies for tool use. J. Neuroscience, 18(23)10105-10115. 2 RSLSteeper, Rochester, United Kingdom.

I01RX001341

2014-01-01

2016-12-31