English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Human activity recognition (HAR) with wearables is one of the serviceabletechnologies in ubiquitous and mobile computing applications. Thesliding-window scheme is widely adopted while suffering from the multi-classwindows problem. As a result, there is a growing focus on joint segmentationand recognition with deep-learning methods, aiming at simultaneously dealingwith HAR and time-series segmentation issues. However, obtaining the fullactivity annotations of wearable data sequences is resource-intensive ortime-consuming, while unsupervised methods yield poor performance. To addressthese challenges, we propose a novel method for joint activity segmentation andrecognition with timestamp supervision, in which only a single annotated sampleis needed in each activity segment. However, the limited information of sparseannotations exacerbates the gap between recognition and segmentation tasks,leading to sub-optimal model performance. Therefore, the prototypes areestimated by class-activation maps to form a sample-to-prototype contrastmodule for well-structured embeddings. Moreover, with the optimal transporttheory, our approach generates the sample-level pseudo-labels that takeadvantage of unlabeled data between timestamp annotations for furtherperformance improvement. Comprehensive experiments on four public HAR datasetsdemonstrate that our model trained with timestamp supervision is superior tothe state-of-the-art weakly-supervised methods and achieves comparableperformance to the fully-supervised approaches.

Timestamp-supervised Wearable-based Activity Segmentation and Recognition with Contrastive Learning and Order-Preserving Optimal Transport