J. Michael Herrmann, Armin Biess, Florentin Wörgötter, and Otto Bock HealthCare GmbH

The project is devoted to the robust and efficient signal analysis of myoelectric data obtained from non-invasive electrode arrays. The goal of the project is the control of prostheses with multiple degrees of freedom and additional sensors based on this data analysis, such that it achieves a flexibility that compares to natural movements.

Advanced prosthetic devices for the upper extremities are usually controlled by myoelectric signals derived from residual or still available muscles. However, conventional transcutaneous recording of movement activity is quite limited in terms of the degrees of freedom that can be controlled reliably and simultaneously. Advances in signal processing and nerve transfer surgery (Kuiken et al. 2007) have considerably improved the achievable signal quality, which now enables some amputees to benefit from recent developments of prosthetic hands. For example, the “Michelangelo” hand (Fig. SP3b) provides proportional control for multi-axial movements, but a complexity reduction had to be achieved by controlling movement synergies rather than the individual degrees of freedom. Even if high-bandwidth signal transduction from the subject to the device is in sight, the information flow in the reverse direction will retain severe limitations. In order to exploit recent advances in robotics in prosthetic devices appropriate feedback signals must be reconstructed from an optimal combination of local sensors and efficient data analysis.



^{Fig.SP3b: The multidimensional myographic signals (left) are transformed into a space of invariant features. Features are combined with sensory input at the device to achieve flexible control of an advanced hand prosthesis (Michelangelo, on the right).}