Long short-term memory networks for body movement estimation

Vlutters, S. (2016) Long short-term memory networks for body movement estimation.

[img]
Preview
PDF
33MB
Abstract:This paper introduces an approach to reconstruct full body motion from a small set of inertial sensors using a long short-term memory (LSTM) network. Although a small set of sensors provides incomplete information for this task, missing degrees of freedom are estimated based on pre-recorded full body motions. Several hyperparameters were tested and their effects on the LSTM’s capabilities of synthesizing full body motion were evaluated and compared with a standard feedforward neural network (FFNN). Our results show that the LSTM performs no better than a FFNN on this problem, indicating that information from the past is unhelpful for estimating missing degrees of freedom. Also, it was found that the networks have difficulties making correct estimations when excluding positional information from our set of sensors as estimator.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Human Media Interaction MSc (60030)
Link to this item:http://purl.utwente.nl/essays/71207
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page