CONTROL STRATEGIES FOR HAND PROSTHESES USING MYOELECTRIC PATTERNS

1 Introduction In Germany 1999 26661 persons were registered who had an amputation of an arm or a complete hand [?]. The main factors for a loss of an upper extremity are accidents followed by general diseases and injuries from war. Although the loss of an upper limb results in a drastic restriction of life quality only few amputees are provided with artificial hands. The main factors for the rejection of conventional prosthetic hands are low functionality and robot-like movement of these artificial limbs. Arm prostheses are usually controlled by myoelectric sensors measuring the electric muscle activities. These sensors detect the level of muscle contraction and give amputees the possibility to control a mechanical prosthesis by muscle activity. The main problems analyzing myoelec-tric patterns are noisy signals so that data cannot be processed reliably. Common myoelectric systems use simple thresholds to detect the state of muscle contraction and to classify two to three movements [?]. Recently, the Institute for Applied Computer Science of the Forschungszentrum Karlsruhe presented an artificial hand with the ability to move all finger joints independently (section 2). To convert this high number of degrees of freedom into movement possibilities new control strategies have to be developed. Several teams are trying to improve a corresponding classification process by using complex preprocessing algorithms and artificial neural networks [?, ?, ?, ?]. This paper presents an fuzzy-based approach to increase movement possibilities of prosthe-ses using algorithms that can be implemented in portable environments (microcontroller). The aims of this paper are