Cheng Yao, Liang Xiaojun, Xu Yi, Kuang Xin
Shaoxing University Yuanpei College, Shaoxing, China.
College of Humanities, Zhaoqing Medical College, Zhaoqing, China.
Front Neurorobot. 2022 Jun 27;16:819784. doi: 10.3389/fnbot.2022.819784. eCollection 2022.
The primary research purpose lies in studying the intelligent detection of movements in basketball training through artificial intelligence (AI) technology. Primarily, the theory of somatosensory gesture recognition is analyzed, which lays a theoretical foundation for research. Then, the collected signal is denoised and normalized to ensure that the obtained signal data will not be distorted. Finally, the four algorithms, decision tree (DT), naive Bayes (NB), support vector machine (SVM), and artificial neural network (ANN), are used to detect the data of athletes' different limb movements and recall. The accuracy of the data is compared and analyzed. Experiments show that the back propagation (BP) ANN algorithm has the best action recognition effect among the four algorithms. In basketball training athletes' upper limb movement detection, the average accuracy rate is close to 93.3%, and the average recall is also immediate to 93.3%. In basketball training athletes' lower limb movement detection, the average accuracy rate is close to 99.4%, and the average recall is immediate to 99.4%. In the detection of movements of upper and lower limbs: the recognition method can efficiently recognize the basketball actions of catching, passing, dribbling, and shooting, the recognition rate is over 95%, and the average accuracy of the four training actions of catching, passing, dribbling, and shooting is close to 98.95%. The intelligent basketball training system studied will help basketball coaches grasp the skilled movements of athletes better to make more efficient training programs and help athletes improve their skill level.
主要研究目的在于通过人工智能（AI）技术研究篮球训练中动作的智能检测。首先，分析体感手势识别理论，为研究奠定理论基础。然后，对采集到的信号进行去噪和归一化处理，以确保所获得的信号数据不会失真。最后，使用决策树（DT）、朴素贝叶斯（NB）、支持向量机（SVM）和人工神经网络（ANN）这四种算法来检测运动员不同肢体动作和召回的数据，并对数据的准确性进行比较和分析。实验表明，反向传播（BP）神经网络算法在这四种算法中具有最佳的动作识别效果。在篮球训练中运动员上肢动作检测方面，平均准确率接近93.3%，平均召回率也接近93.3%。在篮球训练中运动员下肢动作检测方面，平均准确率接近99.4%，平均召回率接近99.4%。在上下肢动作检测中：该识别方法能够高效识别接球、传球、运球和投篮等篮球动作，识别率超过95%，接球、传球、运球和投篮这四项训练动作的平均准确率接近98.95%。所研究的智能篮球训练系统将有助于篮球教练更好地掌握运动员的技术动作，制定更高效的训练计划，并帮助运动员提高其技术水平。
