From General Actions to Domain-Specific Monitoring: Prior-Adaptive Transfer for Skeleton-Based Action Recognition

2026-07-03

AI Summary

This work addresses the performance limitations of directly transferring generic skeleton-based action recognition models to specialized tasks such as healthcare monitoring, where irrelevant motion priors hinder effectiveness. To overcome this, the authors propose the PATS framework, which employs an adaptive prior transfer mechanism that leverages model compression to iteratively prune task-irrelevant spatiotemporal priors while preserving those pertinent to the target domain. Built upon skeleton sequences extracted from videos and a spatiotemporal backbone pretrained on general action recognition, PATS achieves significant improvements over existing baselines in both Alzheimer’s disease detection and fall detection tasks, simultaneously enhancing accuracy and computational efficiency.

Abstract

Skeleton-based action recognition models have recently shown strong performance on large-scale benchmarks with general actions. However, directly transferring them to domain-specific tasks e.g., healthcare monitoring, is often suboptimal, as such tasks are narrow in scope and may be relevant to only a subset of general motion priors. Moreover, not all pretrained motion patterns are equally useful for a specific task, and retaining less relevant components may hinder adaptation and increase computational cost. To address these challenges, we propose Prior-Adaptive Transfer of Skeletons (PATS), a framework that adapts general skeleton-based models by selectively retaining task-relevant motion priors while filtering redundant ones during transfer. PATS follows a standard pipeline that extracts skeleton signals from videos and employs a spatio-temporal backbone pre-trained on general actions. The key contribution lies in a novel Adaptive Prior Transfer module, which performs model compression as a prior selection mechanism through iterative pruning and refinement. Experiments on two specific action recognition tasks, Alzheimer's detection and fall detection, show consistent improvements in both performance and efficiency over competitive baselines. The code will be released upon acceptance.

Problem

• skeleton-based action recognition
• domain-specific monitoring
• transfer learning
• motion priors
• model adaptation

Innovation

• Prior-Adaptive Transfer
• Skeleton-Based Action Recognition
• Model Compression
• Motion Priors
• Domain-Specific Adaptation