摘要
适当的体育运动有利于青少年身体健康,但是大多数青少年在运动过程中,盲目地进行高强度的体育锻炼,很容易造成身体的损伤甚至危及生命。因此,为了实现对青少年运动的合理安排和监测,本研究提出了一种基于长短期记忆人工神经网络(Long Short-Term Memory,LSTM)与心电图(Electrocardiogram,ECG)信号的青少年运动强度识别方法。该方法可以在体育锻炼中实时监测运动强度,防止体育运动中不合理锻炼带来的危险。本研究算法采用多层的LSTM网络提取运动过程中的ECG信号特征,在网络中加入注意力机制,模仿生物的视觉注意力行为,对一段时间序列中的不同区域区别对待,重点关注特征区域,抑制无用信息,进一步提升监测效率和准确率。实验识别准确率可达99.40%,表明所提方法所构建的青少年运动强度诊断模型具有较高的诊断精度,且具有较强的泛化能力。
Appropriate sport is beneficial to the health of teenagers.However,most teenagers blindly engage in highintensity physical exercise which easily causes physical damage or even endanger life.Therefore,in order to realize the reasonable arrangement and monitoring of teenagers exercise intensity,a teenagers exercise intensity recognition method based on LSTM(Long Short-Term Memory artificial neural network)network and ECG(Electrocardiogram)signal was proposed.The method can monitor the exercise intensity in real time during physical exercise,and prevents the danger caused by unreasonable exercise in sport.The algorithm in this study used a multi-layer LSTM network to extract the ECG signal features in the process of motion,added an attention mechanism to the network,imitated the visual attention behavior of the creatures,treated regions in a time series differently,focused on feature regions,and suppress useless information,to improve monitoring efficiency and accuracy.The experimental recognition accuracy of this study can reach 99.40%.It showed that the teenagers exercise intensity diagnostic model constructed with the method in this study has high diagnostic accuracy,as well as high accuracy and generalization ability.
作者
董晋
季炜然
DONG Jin;JI Wei-ran(School of Physical Education,Shanxi University,Taiyuan 030006,China)
出处
《印刷与数字媒体技术研究》
CAS
北大核心
2023年第6期49-58,共10页
Printing and Digital Media Technology Study
基金
山东省社会科学规划研究项目(No.21CTYJ03)。
