Recurrent Neural Network Based Early Prediction of Future Hand Movements

Philipp Koch, Huy Phan, Marco Maass, Fabrice Katzberg, Alfred Mertins

Abstract

This work focuses on a system for hand prostheses thatcan overcome the delay problem introduced by classical approacheswhile being reliable. The proposed approach based on a recurrentneural network enables us to incorporate the sequential nature ofthe surface electromyogram data and the proposed system can beused either for classification or early prediction of hand movements.Especially the latter is a key to a latency free steering of a prosthesis.The experiments conducted on the first three Ninapro databases revealthat the prediction up to200 msahead in the future is possible withouta significant drop in accuracy. Furthermore, for classification, ourproposed approach outperforms the state of the art classifiers eventhough we used significantly shorter windows for feature extraction.
OriginalspracheEnglisch
Seitenumfang4
PublikationsstatusVeröffentlicht - 01.07.2018
Veranstaltung40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society - Hawaii Convention Center, Honolulu, USA / Vereinigte Staaten
Dauer: 17.07.201821.07.2018
Konferenznummer: 141674

Tagung, Konferenz, Kongress

Tagung, Konferenz, Kongress40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society
KurztitelEMBC 2018
Land/GebietUSA / Vereinigte Staaten
OrtHonolulu
Zeitraum17.07.1821.07.18

Fingerprint

Untersuchen Sie die Forschungsthemen von „Recurrent Neural Network Based Early Prediction of Future Hand Movements“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren