Datenbestand vom 09. März 2024

Warenkorb Datenschutzhinweis Dissertationsdruck Dissertationsverlag Institutsreihen     Preisrechner

aktualisiert am 09. März 2024

ISBN 9783843941495

72,00 € inkl. MwSt, zzgl. Versand


978-3-8439-4149-5, Reihe Informatik

Peter Hegen
Integrated Hardware & Software Platform with Automated FIR Filter Coefficient Adaptation for Advanced Control of Modern Hand Prostheses

182 Seiten, Dissertation Universität der Bundeswehr München (2018), Softcover, A5

Zusammenfassung / Abstract

Current hand prostheses for under-elbow amputees are still based on a simple control scheme, which was introduced decades ago. Based on electromyographic signals recorded from residual muscles, a user can control an opening and a closing movement of the prosthesis by muscle contraction. This simple scheme has proliferated due to its robustness against slow physiological changes compared to newer schemes. This method is however limited in its functionality and requires the user to memorize the current state of the hand prosthesis.

Consequently, research focuses on classification based schemes that are easier to use and allow for a higher number of movement functions. These classification schemes rely on features, which require preconditioning/filtering of the input signals to limit the influence of external noise sources. However, the filters chosen in previous works use fixed parameters to ensure that only frequencies with relevant information are considered. This approach fails to account for the different signal characteristics of individuals and the frequency dependency of each feature, which may result in degraded classification performance.

To address this problem, a novel method is proposed to adapt filters to combinations of features and individual persons, catering to the different signal characteristics and frequency dependencies. Specifically, a constrained optimization algorithm is used to iteratively modify the filter coefficients and optimize the discrimination power of feature combinations.

For evaluating the filter adaptation approach, a platform for research on myographic signals was developed, comprising active sensors, a sleeve to record myographic signals from able-bodied probands and a commercial prosthesis. Custom electronics and firmware were developed for the prosthesis to provide additional functions, like the concurrent support for analog and digital sensors and user-changeable filters. Complementary custom applications for mobile and stationary devices provide support for different aspects: (1) the prosthesis can be configured through a Bluetooth-connected device; (2) recorded signals can be analyzed to derive suitable parameters for the classifier and the filters used in the prosthesis; and (3) recording sessions are assisted by guiding the proband through the required steps, ensuring proper timing to enable the automatic labeling of the recorded signals for later classification.