Perceptual Feature Integration for Sports Dancing Action Scenery Detection and Optimization
Authors:
Open Access
- 2 September 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 122101-122113
Abstract
Deciphering the complex semantics within varied dancing sceneries is crucial for a multitude of AI endeavors. It can facilitate applications like dancing action optimization and dancing education. In our research, we propose a sophisticated approach to discerning multi-faceted perceptual visual features for accurately identifying dancing scenic imagery with intricate spatial designs. Our work centers on crafting a deep hierarchical structure adept at simulating human gaze patterns, utilizing the BING metric to pinpoint objects and their components within scenes at different scales. To emulate human visual dynamics, we introduce a Robust Deep Active Learning (RDAL) methodology, systematically creating gaze shift paths (GSPs) and capturing their profound representations. A key novelty of RDAL is its resilience to inaccuracies in labeling, employing a strategically designed sparse penalty framework that facilitates the exclusion of non-informative or irrelevant deep GSP attributes. Furthermore, we propose a manifold-regularized feature selector (MRFS) to isolate premium deep GSP features, concurrently developing a linear SVM for dancing scene recognition. Our method’s efficacy, validated through rigorous testing, not only showcased its enhanced performance across conventional scenic datasets but also highlighted the exceptional discriminating power of deep GSP features in a specialized dataset for recognizing different dancing actions. Finally, the dancing actions can be optimized using a probabilistic model.Keywords
Funding Information
- Hunan Province General Higher Education Teaching Reform Research Project (HNJG-20231319)
- Scientific Research Project of Hunan Provincial Department of Education (23C0405)
This publication has 66 references indexed in Scilit:
- HFA-GTNet: Hierarchical Fusion Adaptive Graph Transformer network for dance action recognitionJournal of Visual Communication and Image Representation, 2023
- Dance Video Motion Recognition Based on Computer Vision and Image ProcessingApplied Artificial Intelligence, 2023
- An Automated Method for Identification of Key frames in Bharatanatyam Dance VideosIEEE Access, 2021
- Algorithms for semantic segmentation of multispectral remote sensing imagery using deep learningISPRS Journal of Photogrammetry and Remote Sensing, 2018
- Weakly supervised vehicle detection in satellite images via multi-instance discriminative learningPattern Recognition, 2017
- Unsupervised Learning of Semantics of Object Detections for Scene Categorization2014
- Selective Search for Object RecognitionInternational Journal of Computer Vision, 2013
- A Hierarchical and Contextual Model for Aerial Image ParsingInternational Journal of Computer Vision, 2009
- A tutorial on spectral clusteringStatistics and Computing, 2007
- What attributes guide the deployment of visual attention and how do they do it?Nature Reviews Neuroscience, 2004