Abstract: A myoelectric pattern that is an action potential generated by a coordinated action of a plurality of muscles is measured by a group of surface electrodes on a skin surface, and a feature pattern is extracted from the measured myoelectric pattern. Redundant coding is used to encode the extracted feature pattern into a bit pattern, the coded bit pattern is subjected to pattern classification, and output control signal is generated.

Abstract: A feature-pattern extraction apparatus uses a logarithmic transformation processing to extract a feature pattern from a myoelectric pattern that is a muscle action potential. The extracted feature patterns are classified, and an output control signal is generated. By using the logarithmic transformation processing to extract feature patterns, distribution bias is reduced, improving classification accuracy.

Abstract: An artificial hand is constructed to sufficiently rotate the fingers as a unit using a smaller small-sized motor by decreasing the magnitude of the radius of rotation of its rotating motion. The artificial hand has a support part for rotatably supporting the fingers; a conversion device for converting a rotating output of a driving device to a linear motion; and a finger opening-and-closing operation part for converting a linear motion of the conversion device to an opening-and-closing operation of the fingers. A holding part rotatably holds the support part above or at an upper part of the driving device; and another driving device is arranged in a terminal device held by a wrist part. A rotating output of another driving device is converted to a rotational motion of the support part; and a rotating mechanism rotates the fingers and driving device as a unit around the holding part.

Abstract: An object of the present invention is to provide an artificial hand which can sufficiently rotate the whole fingers using a smaller small-sized motor by decreasing the magnitude of the radius of rotation of its rotating motion.

Abstract: A myoelectric pattern that is an action potential generated by a coordinated action of a plurality of muscles is measured by a group of surface electrodes on a skin surface, and a feature pattern is extracted from the measured myoelectric pattern. Redundant coding is used to encode the extracted feature pattern into a bit pattern, the coded bit pattern is subjected to pattern classification, and output control signal is generated.

Abstract: A feature-pattern extraction apparatus uses a logarithmic transformation processing to extract a feature pattern from a myoelectric pattern that is a muscle action potential. The extracted feature patterns are classified, and an output control signal is generated. By using the logarithmic transformation processing to extract feature patterns, distribution bias is reduced, improving classification accuracy.