Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition.

Ordóñez FJ, Roggen D. · 2016

Read Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition. by Ordóñez FJ, Roggen D. (2016) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4732148.txt.

Description

Francisco Javier Ordóñez and Daniel Roggen built a deep learning system that recognizes human activities from wearable sensors, in Sensors in 2016. Activity recognition usually depends on hand-designed features extracted from sensor streams, a craft bottleneck. Their framework pairs convolutional layers, which learn features from raw signals, with recurrent LSTM layers. The recurrent stage tracks how movements unfold over time. It needs no expert feature design and fuses sensor types naturally.

On two datasets, one from a public challenge, the system beat competing deep networks lacking the recurrent stage, by four percent on average. Tests with single sensor types also worked, though fusion helped. The authors probed which design settings matter most, offering tuning guidance. For assistive technology, from fall detection to rehabilitation support, automatic feature learning lowers the cost of each new application.

Catalog record

CREATOROrdóñez FJ, Roggen D.
YEAR2016
COLLECTIONPAPERS
FORMATTXT
FILE SIZE70 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4732148

Browse the full library · About the Tower of Babel