Myoelectric Control of Artificial Limbs – Is There a Need to Change Focus?
By Ning Jiang, Strahinja Dosen, Klaus-Robert Müller, and Dario Farina
NOTE: This is an overview of the entire article, which appeared in the September 2012 issue of the IEEE Signal Processing magazine.
Click here to read the entire article.
This article reviews the concept of myoelectric control (principally of the upper limbs) and the state of the art in industy and academia. The authors assert that there is a gap between industrial and academic achievements and that this gap will continue to expand unless a change of focus in systems for myoelectric control occurs. Their focus is on the challenges facing myoelectric control and on possible approaches to bridge the gap.
The first myoelectric controllers (from the 1950s) used surface EMG signals This approach is still used by the vast majority of commercially available powered prostheses. Since the 1950’s, researchers have focused on pattern classification techniques, where different types of muscle activation signals are extracted from the EMG. High accuracies (>90%) have been reported. However, this improved accuracy does not seem to significantly improve the usability of the artificial upper limb, or its acceptance by patients.
The authors discuss four factors that are missing from conventional controllers, which they believe are limiting the usefulness of prostheses:
-
The pattern classification techniques do not provide continuous and proportional control of the multiple degrees-of-freedom (DOF) that are associated with joints. Adding proportionality into the pattern classification technique impairs its accuracy, so these techniques do not lend themselves to proportional control, especially across multiple DOFs.
-
There is no use of sensory feedback of limb position in most prostheses. Such feedback is an integral part of the control systems of natural limbs. But recent research on sensory feedback in myoelectric control has been negligible, compared to the resources devoted to decoding motor actions.
-
There is no adaptability mechanism to respond to changes in EMG signals, although received EMG signals do change due to fatigue, displacement of electrodes, etc. The authors point out that adaptive strategies have been implemented in other rehabilitation technologies, but applications in prosthesis control have been very limited.
-
Other sensor modalities than EMG have not been used. They propose, for instance, using inertial sensors to measure the orientation and movement of a limb. This could lead to controllers that can operate more autonomously, without the need for conscious attention on the part of the user.
The authors express their intention to raise the awareness of the need for a change of focus in research efforts in order to make artificial prostheses more useful to the general amputee population.
ABOUT THE AUTHORS
Ning Jiang (ning.jiang@ottobock.de) is a Marie Curie Fellow with Strategic Technology Management, Otto Bock Healthcare GmbH and is also affiliated with the Department of Neurorehabilitation Engineering, University Medical Center, Göttingen, Germany.
Strahinja Dosen (strahinja.dosen@bccn.uni-goettingen.de) is a research scientist at the Department of Neurorehabilitation Engineering, University Medical Center, Göttingen, Germany.
Klaus-Robert Müller (klaus-robert.mueller@tu-berlin.de) is a full professor and head of the Machine Learning Group, Technical University of Berlin, Berlin, Germany, and is also affiliated with the Department of Brain and Cognitive Engineering, Korea University, Anam-dong, Seongbuk-gu, Seoul 136-713, Korea.
Dario Farina (dario.farina@bccn.uni-goettingen.de) is a full professor and head of the Department of Neurorehabilitation Engineering, University Medical Center, Göttingen, Germany.
