Research on an Application Model of Action Recognition Technology in Human-Computer Interaction

View at publisher Summarise with VIRAT
Authors: Hongbing Ba, Xi Wu
Cited by: 0
References: 1

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

To address the feedback lag and reliance on subjective judgment in traditional human-computer interaction during sports training, this study designed a simplified three-layer "perception-processing-interaction" action recognition human-computer interaction model, adapted for middle- and long-distance running posture correction and basketball shooting training scenarios. The perception layer uses an RGB camera (1280×720 resolution) and a 6-axis MEMS inertial sensor (±16g acceleration range) for multimodal data acquisition. The processing layer, based on an optimized MobileNetV3-Small model (without the SE module), simplifies the "preprocessing-keyframe extraction-matching" process. The interaction layer establishes a "recognition result-correction instruction" mapping mechanism, integrating tactile and visual feedback. Experimental verification showed that after 30-35 training cycles in two scenarios, the model achieved stable accuracy (88.5% ± 0.3% for middle-distance running and 86.3% ± 0.5% for basketball), interaction latency ≤ 98ms (highest for shooting), and athlete and coach satisfaction ≥ 4.1 points (feedback clarity rating reached 4.5 points). Simulation plots show that accuracy increases with training cycles, latency increases slightly with movement complexity, and report generation time increases linearly with training duration (generating a report for 50 minutes of training data takes 12.5 seconds). This research confirms that the model can meet the low-cost, high-real-time interaction requirements of sports scenarios, providing technical support for intelligent training.

Fields of Study & Subjects

Human Pose and Action Recognition Context-Aware Activity Recognition Systems Hand Gesture Recognition Systems

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1444–1448
Publication date
28 January 2026
Language
EN