摘要
针对青少年耐力素质逐年下降和缺乏具有便携性与监督性的训练系统的现状,提出一种基于注意力(Attention)机制与门控循环单元(GRU)的、面向耐力训练的AttentionGRU动作识别方法。该方法包括数据采集、数据处理和动作识别,在手机端实现了对耐力训练动作的识别。基于新构建的耐力训练动作数据集,将本文提出的方法与常用动作识别算法——长短期记忆(LSTM)、GRU进行实验对比。实验结果表明:本文提出的方法表现更佳,准确率达到99.56%,较LSTM高出1.5个百分点。
Aiming at the status that the endurance quality of adolescents is declining year by year and the lack of a portable and supervised training system,an Attention-GRU action recognition methods based on Attention mechanism and gated recurrent unit(GRU)is proposed for endurance training.The method includes data acquisition,data processing and action recognition,and the recognition of endurance training actions is realized on the mobile phone.Based on the newly constructed endurance training action dataset,the proposed method is experimentally compared with the commonly used action recognition algorithms,long short-term memory(LSTM)and GRU.The experimental results show that the proposed method performs better,with an accuracy rate of 99.56%,which is 1.5 percentage points higher than that of LSTM.
作者
田新壮
孙少明
王君洪
TIAN Xinzhuang;SUN Shaoming;WANG Junhong(Institutes of Physical Science and Information Technology,Anhui University,Hefei 230601,China;Hefei Institutes of Physical Science,Chinese Academy of Sciences,Hefei 230031,China;CAS(Hefei)Institute of Technology Innovation,Hefei 230088,China)
出处
《传感器与微系统》
CSCD
北大核心
2024年第3期120-124,共5页
Transducer and Microsystem Technologies
基金
国家重点研发计划资助项目(2018YFC2001304)。
关键词
训练监督
动作识别
骨骼点识别
神经网络
training supervision
action recognition
bone point recognition
neural network
