D8.3 - EMG Based Input and Control System for Lower Limb Prostheses
- Event
- SENSOR+TEST Conferences 2011
2011-06-07 - 2011-06-09
Nürnberg - Band
- Proceedings SENSOR 2011
- Chapter
- D8 - Medical III
- Author(s)
- H. von Rosenberg, F. Dennerlein - Fraunhofer-Institut (IPA), Stuttgart (Germany)
- Pages
- 664 - 668
- DOI
- 10.5162/sensor11/d8.3
- ISBN
- 978-3-9810993-9-3
- Price
- free
Abstract
We present a voluntary control input system for lower limb prostheses. For this purpose we use electromyography (EMG) with skin surface electrodes which is a common method for non-invasive measurement of electrical potential during muscle contraction but has not yet been applied successfully to lower limb prostheses due to the difficult signal conditions within the prosthesis socket. Our system will be integrated in the prosthesis socket, making the control system comfortable to wear and easy to use as no electrodes or other parts must be glued to the skin or manually repositioned and placed each day the prosthesis is worn. The control system is designed to allow the user to control a special function within an active powered prosthesis system, e.g. lifting the foot during walking or climbing stairs which are elementary tasks, performed by the subconscious mind of an non-amputee.
The control signal is derived by using remaining muscles of the residual limb. The active prosthesis function can be triggered either by a distinct stimulation of the muscle which is observed or by the subconscious muscle stimulation remaining after the amputation.
Our solution uses an EMG sensor array with direct skin contact. A virtual EMG signal is generated that is superior to each single channel representing the best EMG muscle reading, even if the skin and surface is moving below the sensor array. Signal artifacts and disturbances are reduced in this virtual EMG system and the subsequent signal processing based on a set of pattern-recognition algorithms is reduced in complexity as the input signal is minimized to a single sensor signal leading to less demand on computational power. This allows sophisticated applications in embedded systems which can directly be integrated into a prosthesis system.
The current motion is also captured to allow proper detection of the user's voluntary control signal. The motion detection logic alters the EMG pattern recognition system, switching between different feature sets. This motion classification is also performed in real-time and covers typical situations, in which a voluntary prosthesis control signal is of high importance.
Our development also offers a potential for further applications like computer input devices, security and safety application, sports and gaming or consumer device control and is not limited to medical applications.