基于MSRC-BiGRU-SA的人体活动识别  

Human Activity Recognition Based on MSRC-BiGRU-SA

作　　者：芦平 于增辉 华国环[1] LU Ping;YU Zeng-hui;HUA Guo-huan(College of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;Zhongke Nanjing Intelligent Technology Research Institute,Nanjing 211135,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,江苏南京210044 [2]中科南京智能技术研究院,江苏南京211135

出　　处：《中国电子科学研究院学报》2025年第1期25-32,共8页Journal of China Academy of Electronics and Information Technology

基　　金：国家重点研发计划(2022YFB4401300)。

摘　　要：针对目前基于可穿戴传感器的复杂人体活动分类算法大多忽略对多尺度特征的提取和关键特征捕捉的问题,文中提出一种多尺度残差卷积网络叠加双向门控循环单元和自注意力机制(MSRC-BiGRU-SA)的模型。首先,通过MSRC模块充分提取传感器数据的多尺度空间和时间特征并有效融合原始数据的特征信息,增强特征的表达能力和鲁棒性;其次,利用BiGRU模块充分捕捉时间序列的前后依赖关系;最后,通过SA模块增强模型对复杂活动关键特征的捕捉能力以提升分类性能。实验结果表明,在公开数据集上,该模型对复杂活动的分类准确率达到97.50%,相较于原始CNN-BiGRU模型提升了5.77%,与现有先进模型相比,具有更好的识别效果。In response to the problem that most of the current complex human activity classification algorithms based on wearable sensors ignore the extraction of multi-scale features and the capture of key features,this paper proposes a model of multi-scale residual convolutional network superimposed with bidirectional gated recurrent units and self-attention mechanism(MSRC-BiGRU-SA).First,the MSRC module is used to fully extract multi-scale spatial and temporal features from sensor data and effectively fuse the feature information of the original data,so as to enhance the expression ability and robustness of the features.Second,the BiGRU module is utilized to fully capture the backward and forward dependencies of the time series.Finally,the SA module is used to enhance the model’s ability to capture key features of complex activities to improve the classification performance.The experimental results show that the model achieves 97.50%classification accuracy for complex activities on the public dataset,which is an improvement of 5.77%compared to the original CNN-BiGRU model,and has better recognition results compared to the existing advanced models.

关 键 词：复杂人体活动识别 卷积神经网络 双向门控循环单元 可穿戴传感器 深度学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]