Pose-critical keypoint attention model for dynamic yoga pose classification

Yoga pose detection and classification have garnered considerable attention in recent years due to their significant applications in various domains, including fitness, health monitoring, and rehabilitation programs. Accurate detection and classification of yoga poses not only enhance the effectiveness of fitness regimes but also facilitate the assessment of practitioners’ health and recovery processes. However, existing methods often struggle with the challenges posed by high-dimensional data processing, particularly in real-time applications. This paper introduces the Pose-Critical Keypoint Attention (PCKA) model, a novel approach designed to overcome the limitations of traditional methods in yoga pose analysis. The PCKA model employs an advanced attention mechanism to dynamically prioritize the most relevant keypoints from a comprehensive set of 33 extracted keypoints. This dynamic focus allows for efficient classification without the drawbacks associated with static keypoint reduction, ensuring that the model remains adaptable and robust across various poses. Extensive experiments evaluate the performance of the PCKA model against conventional classification frameworks. The results indicate that this model significantly outperforms traditional approaches, achieving an impressive accuracy of 91.08%, precision of 89.70%, recall of 91.93%, and an F1-score of 90.78%. These metrics demonstrate the model’s capability to deliver accurate pose classification while providing real-time feedback for corrective measures, which is crucial for users striving to improve their yoga practices.