Our research is directed towards developing a hand gesture recognition system that can predict hand movements from multisensory biosignal data. Thus, the focus of the hand movement prediction task is on data from surface electromyography sensors and data from an inertial measurement unit. Especially, recurrent neural networks (RNNs) can help to achieve good performance for the regression and classification task while requiring only a short processing time and a reasonable computational cost. The development of RNNs with short delay times and low computational complexity are of utmost importance in the research project to give neural networks a realistic chance to be used in smart prosthetics in the future.

Team & Contact

• Philipp Koch (ph.koch@uni-luebeck.de)
• Mark Dreier
• Marco Maaß
• Huy Phan
• Ole Thomsen
• Prof. Dr. Alfred Mertins

Papers

1. Koch, P., Dreier, M., Maass, M., Phan, H. and Mertins, A.: RNN With Stacked Architecture for sEMG based Sequence-to-Sequence Hand Gesture Recognition in 28th European Signal Processing Conference (EUSIPCO), pp. 1600-1604, 2020
2. Koch, P., Dreier, M., Larsen, A., Parbs, T. J., Maass, M., Phan, H. and Mertins, A.: Regression of Hand Movements from sEMG Data with Recurrent Neural Networks in 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pp. 3783--3787, 2020