High-Level Features for Human Activity Recognition and Modeling
by , ,
Abstract:
High-Level Features (HLF) are a novel way of describing and processing human activities. Each feature captures an interpretable aspect of activities, and a unique combination of HLFs defines an activity. In this article, we propose and evaluate a concise set of six HLFs on and across the CSL-SHARE and UniMiB SHAR datasets, showing that HLFs can be successfully extracted with machine learning methods and that in this HLF-space activities can be classified across datasets as well as in imbalanced and few-shot learning settings. Furthermore, we illustrate how classification errors can be attributed to specific HLF extractors. In person-independent 5-fold cross-validations, the proposed HLFs are extracted from 68% up to 99% balanced accuracy, and activity classification achieves 89.7% (CSL-SHARE) and 67.3% (UniMiB SHAR) accuracy. Imbalanced and few-shot learning results are promising, with the latter converging quickly. In a person-dependent evaluation across both datasets, 78% accuracy is achieved. These results demonstrate the possibilities and advantages of the proposed high-level, extensible, and interpretable feature space.
Download PDF
PDF URL: https://www.csl.uni-bremen.de/cms/images/documents/publications/HartmannLiuSchultz_Springer2023.pdf
Reference:
High-Level Features for Human Activity Recognition and Modeling (Yale Hartmann, Hui Liu, Tanja Schultz), In Biomedical Engineering Systems and Technologies (Ana Cecília A. Roque, Denis Gracanin, Ronny Lorenz, Athanasios Tsanas, Nathalie Bier, Ana Fred, Hugo Gamboa, eds.), Springer Nature Switzerland, 2023.
Bibtex Entry:
@inproceedings{hartmann2023har_high_level_features,
  title = {High-Level Features for Human Activity Recognition and Modeling},
  author = {Hartmann, Yale and Liu, Hui and Schultz, Tanja},
  booktitle = {Biomedical Engineering Systems and Technologies},
  editor = {Ana Cec{\'i}lia A. Roque and Denis Gracanin and Ronny Lorenz and Athanasios Tsanas and Nathalie Bier and Ana Fred and Hugo Gamboa},
  year = {2023},
  publisher = {Springer Nature Switzerland},
  address = {Cham},
  pages= {141--163},
  isbn = {978-3-031-38854-5},
  doi = {10.1007/978-3-031-38854-5_8},
  url = {https://www.csl.uni-bremen.de/cms/images/documents/publications/HartmannLiuSchultz_Springer2023.pdf},
  abstract = {High-Level Features (HLF) are a novel way of describing and processing human activities. Each feature captures an interpretable aspect of activities, and a unique combination of HLFs defines an activity. In this article, we propose and evaluate a concise set of six HLFs on and across the CSL-SHARE and UniMiB SHAR datasets, showing that HLFs can be successfully extracted with machine learning methods and that in this HLF-space activities can be classified across datasets as well as in imbalanced and few-shot learning settings. Furthermore, we illustrate how classification errors can be attributed to specific HLF extractors. In person-independent 5-fold cross-validations, the proposed HLFs are extracted from 68{\%} up to 99{\%} balanced accuracy, and activity classification achieves 89.7{\%} (CSL-SHARE) and 67.3{\%} (UniMiB SHAR) accuracy. Imbalanced and few-shot learning results are promising, with the latter converging quickly. In a person-dependent evaluation across both datasets, 78{\%} accuracy is achieved. These results demonstrate the possibilities and advantages of the proposed high-level, extensible, and interpretable feature space.},
}