Guidance Method for Specialized Strength Training of Curling Athletes Based on Action Matching

When analyzing and comparing sports data, the description of human body structure changes with the perspective, and there are significant differences in the representation of bones in the same posture captured from different perspectives. This change can lead to poor performance of curling athletes in specific strength training guidance programs. To enhance the effectiveness of specialized strength training for curling athletes, this paper introduces a novel approach. Specifically, it proposes a motion matching-based method to guide and inform their specialized strength training, offering valuable insights and suggestions tailored to their unique needs. First, Kinect v2 is used to collect training motion data of curling athletes, and the collected training motion data are used to construct three-dimensional coordinates of human skeletal joint points. The process involves several steps. First, the three-dimensional coordinates of curling athletes’ human joint points are normalized to produce standardized athlete training motion data. Second, to ensure the analysis focuses solely on individual postures without being influenced by the continuity of human movement, the posture is examined one by one, while the semantic relationship between consecutive actions is intentionally overlooked. A bidirectional LSTM network is utilized to extract the embedded features from the sequence of specialized strength training actions. This step facilitates the guidance and processing of curling athletes’ specialized strength training. The dynamic time warping (DTW) algorithm is employed to extract three-dimensional human skeleton data. These data serve as the foundation for comparing and matching the athletes’ movements. The Cosine DTW algorithm is applied to compute the directional difference between the two force training action vectors that need to be matched. This calculation helps in identifying the similarity or dissimilarity in the direction of the athletes’ movements. When evaluating the overall similarity of the skeleton, the Cosine DTW distances of each skeleton feature vector are summed up. The magnitude of this accumulated value provides an indication of the overall similarity of the skeleton in that particular frame. This step completes the feature matching of curling athletes’ specialized strength training movements. The experimental results show that both compound training and traditional strength training can improve athletes’ strength quality. The training effect of compound training is more obvious, and it can also enhance athletes’ coordination ability, balance ability, and physical stability ability, and improve their specialized sports performance. Therefore, it is also suitable as a strength training method for curling athletes. The nutritional intake of athletes in specialized strength training should be supplemented according to their training intensity and physical condition, and balanced nutritional supplementation is a prerequisite for ensuring effective training.