Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

BCCN AG-Seminar

Tuesday, 13.03.2007 16 c.t.

Adaptive Myo-Signal Processing for Multidimensional Prosthesis Control

by Dr. Horst Willburger

Contact person: J. Michael Herrmann


Seminarraum Haus 2, 4. Stock (Bunsenstr.)


Beside body-powered prostheses, electrically driven prosthetic hands and arms are of immense help in everyday activities for many individuals with amputations or congenitally deficient upper limbs. Surface Electromyographic (sEMG) signals are key sources for retrieving control information since they are easily acquired without the need for surgery. A possible approach to control multidimensional hand prosthesis is to use the remaining motility of the amputee to form commands. Electrodes located at the amputees stump pick up the associated sEMG signal. Characteristic features are extracted by a signal processing stage and then distinguished by a classifier. Our work mainly focuses on improving the classification accuracy by highlighting three basic entities, the sEMG signal pick up, the assembly of the feature set and the selection of the classifier. A new class of features received from closely spaced electrodes are developed showing promising results. These features mainly base on action potential propagation characteristics and show good results, especially for low contraction forces where other features fail. We use support vector machines to analyse and classify the features. As a result of this work, a prosthesis control scheme with surprisingly good user acceptance was developed. The mental effort for the user was highly reduced.

back to overview