Cross-subject and cross-modal transfer for generalized abnormal gait pattern recognition
- Submitting institution
-
University of Glasgow
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11-12026
- Type
- D - Journal article
- DOI
-
10.1109/TNNLS.2020.3009448
- Title of journal
- IEEE Transactions on Neural Networks and Learning Systems
- Article number
- -
- First page
- 546
- Volume
- 32
- Issue
- 2
- ISSN
- 2162-237X
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2020
- URL
-
http://eprints.gla.ac.uk/220329/
- 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
-
4
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- ORIGINALITY: Enables the disentanglement of pattern-specific and subject-specific (biometrics) data representations with novel Deep Neural Networks architectures. It is the first study to successfully transfer knowledge across different sensing modalities of human motion, while also enabling motion re-targeting. SIGNIFICANCE: Introduces approaches to analyse wearable/ambient sensing data and develops brain computer interfaces which are robust to ambiguous representations and low signal-to-noise ratio. RIGOUR: It has been implemented and validated on synchronous acquisitions of complex multi-modal data from 18 healthy volunteers: i) multi-camera tracking, ii)single RGBD data, iii)EMG. A detailed ablation study and comparison with state-of-the-art approaches is presented.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -