Deep, Convolutional, and Recurrent Models for Human Activity Recognition using Wearables
- Submitting institution
-
University of Newcastle upon Tyne
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 224219-115844-1292
- Type
- E - Conference contribution
- DOI
-
-
- Title of conference / published proceedings
- Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence
- First page
- 1533
- Volume
- -
- Issue
- -
- ISSN
- 1045-0823
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2016
- URL
-
http://www.ijcai.org/Abstract/16/220
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
2
- Research group(s)
-
C - Open Lab
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This is an extensively referenced work presented at a prestigious conference. It provides the first unbiased and systematic exploration of the performance of state-of-the-art deep learning approaches on three different recognition problems typical for Human Activity Recognition (HAR) in ubiquitous computing scenarios. The paper introduces a novel approach to regularisation for recurrent networks. Over 4,000+ experiments were used to investigate suitability of each model for different HAR tasks. Impacts on performance of each model’s hyper-parameters are reported and guidelines for use of deep learning techniques in ubicomp applications are presented.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -