基于CBAM-DCNN-BiLSTM的蹴球动作识别与评估

Cuqiu motion recognition and skill assessment based on CBAM-DCNN-BiLSTM

当前蹴球训练方式主要依赖于教练员个人经验,缺乏数据化的科学训练手段.利用惯性传感器对蹴球运动数据进行采集和挖掘,可对运动动作进行有效识别和评估.然而,现有人体活动识别算法模型对下半身运动传感数据的信息敏感度偏低,重要特征表达能力不强,且对时间信息的利用不够充分.基于此,提出了一种由空洞卷积神经网络(DCNN)与双向长短期记忆循环神经网络(BiLSTM)融合的模型,通过将通道注意力机制与空间注意力机制相结合的卷积注意力模块CBAM引入网络中,以提高模型对于重要特征的表达能力.实验结果表明:该模型在蹴球动作识别与评估任务中比同类神经网络模型具有更好的性能,其精确率、召回率以及F1值可分别达到99.05%、99.04%、99.04%.

At present,the way of Cuqiu training mainly depends on coaches'personal experience and lacks of scientific training means.Using inertial sensors to collect and mine the motion data of Cuqiu can effectively identify and evaluate the motion activity.However,the existing human activity recognition algorithm models have low sensitivity to the lower body motion sensing data,weak expression ability of important features,and insufficient utilization of time information.Based on this,this paper proposes a fusion model of dilated convolution neural network(DCNN)and bidirectional long short-term memory recurrent neural network(BiLSTM).The convolutional attention module CBAM,which combines channel attention mechanism and spatial attention mechanism,is introduced into the network to improve the expression ability of the model for important features.The experimental results show that the model has better performance than the similar neural network model in the task of Cuqiu motion recognition and evaluation.It's precision,recall and F1-score can reach 99.05%,99.04%and 99.04%respectively.