计算机科学

计算机科学 ›› 2023, Vol. 50 ›› Issue (8): 193-201.doi: 10.11896/jsjkx.220900124

• 人工智能 • 上一篇    下一篇

基于元学习和注意力机制的用户行为识别研究

王佳昊1, 钟鑫1, 李文雄1, 赵德鑫2   

  1. 1 电子科技大学信息与软件工程学院 成都 610051
    2 军事科学院国防科技创新研究院 北京 100071
  • 收稿日期:2022-09-14 修回日期:2022-12-08 出版日期:2023-08-15 发布日期:2023-08-02
  • 通讯作者: 赵德鑫(zhaodx2008@163.com)
  • 作者简介:(wangjh@uestc.edu.cn)
  • 基金资助:
    电子科技大学-智小金-智能家居联合研究中心项目(H04W210180);内江市科技孵化和成果转化专项资金(2021KJFH004);四川省科技计划项目(2022YFG0212)

Human Activity Recognition with Meta-learning and Attention

WANG Jiahao1, ZHONG Xin1, LI Wenxiong1, ZHAO Dexin2   

  1. 1 School of Information and Software Engineering,University of Electronic Science and Technology,Chengdu 610051,China
    2 National Innovation Institute of Defense Technology,Academy of Military Sciences,Beijing 100071,China
  • Received:2022-09-14 Revised:2022-12-08 Online:2023-08-15 Published:2023-08-02
  • About author:WANG Jiahao,born in 1978,Ph.D,associate professor,is a member of China Computer Federation.His main research interests include IoT,information security and data mining.
    ZHAO Dexin,born in 1984,Ph.D,asso-ciate researcher.His main research interests include underwater IoT,intelligent recognition and autonomous underwater vehicles.
  • Supported by:
    UESTC-ZHIXIAOJING Joint Research Center of Smart Home(H04W210180), Neijiang Technology Incubation and Transformation Funds(2021KJFH004) andScience and Technology Program of Sichuan Province,China(2022YFG0212).

PDF (PC)

767

摘要： 深度学习技术研究的深入,极大地促进了其在行为识别领域的应用和发展。目前基于深度学习的行为识别研究不可避免地依赖于大量的训练数据,而基于传感器数据的行为识别问题往往在实际应用中需要面向不同的新用户,导致存在用户数据个性化的问题且难以解决,即不同个体在进行同一行为动作时不可避免地会产生一些数据差异,模型在面对新用户时并不能保证对其具有良好的预期行为识别度,而每次针对新用户采集大量训练数据以进行重新建模缺乏实施可行性。针对这一难题,小样本学习技术在新的任务上仅使用少量数据就能够达到较好效果,即在行为识别问题上,每个新用户仅需采样少量的数据即可完成分类。文中结合小样本学习和行为识别算法,提出了新的解决方案——MAML-M模型。首先采用基于优化的元学习方法根据用户类型对数据集进行划分,并将其构建为多个用户任务用于训练和测试;然后在MAML-M模型中引入了MAML方法以及基于注意力机制的Memory模块;最后提高模型网络提取并归纳数据特征的能力。通过在MEx数据集上进行对比实验,结果证明在小样本设定下,所提出的MAML-M模型优于传统的深度学习方法。

关键词: 人体行为识别, 小样本学习, 元学习, 注意力机制

Abstract: With the in-depth research of deep learning technology,its application and development in the field of behavior recognition have been greatly promoted.Current research on behavior recognition based on deep learning usually requires a large training data set.But when facing practical applications,new users will inevitably run into personalization issues.This means that even while performing the same activity,different people may use training data sets differently.Existing solutions cannot guarantee to achieve the expected accuracy when dealing with new users.Besides,these models would also be impractical to deploy when gathe-ring training data for new users.Facing this problem,small-sample learning can achieve better results by using only a small number of samples.This means that in the behavior recognition problem,each new user can be classified using a little training data.In this paper,a MAML-M model is proposed by combining few-shot learning and behavior recognition algorithms.Firstly,an optimization-based meta-learning method is adopted to divide the dataset according to users and construct multiple user tasks for trai-ning and testing.Meanwhile,the MAML method and the memory module based on the attention mechanism are introduced into the MAML-M model,which finally improves the ability of the model network to extract and summarize data features.Through experiment on MEx dataset,the proposed MAML-M model shows better performances under small sample sets.

Key words: Human behavior recognition, Small sample learning, Meta-learning, Attention mechanism

中图分类号: 

  • TP181
[1]SABOOR A,KASK T,KUUSIK A,et al.Latest researchtrends in gait analysis using wearable sens-ors and machine learning:A systematic review[J].IEEE Access,2020,8:167830-167864.
[2]SHIH C S,CHOU J,LIN K J.WuKong:Secure Run-Time environment and data-driven IoT applications for Smart Cities and Smart Buildings[J].Journal of Internet Services and Information Security,2018,8(2):1-17.
[3]ZHANG S,LI Y,ZHANG S,et al.Deep Learning in HumanActivity Recognition with Wearable Sensors:A Review on Advances[J].Sensors,2022,22(4):1476.
[4]DANG L M,MIN K,WANG H,et al.Sensor-bas ed and vision-based human activity recognition:A comprehensive survey[J].Pattern Recognition,2020,108:107561.
[5]MEKRUKSAVANICH S,JITPATTANAKUL A.LSTM networks using smartphone data for sensor-based human activity recognition in smart homes[J].Sensors,2021,21(5):1636.
[6]JOBANPUTRA C,BAVISHI J,DOSHI N.Human activity re-cognition:A survey[J].Procedia Computer Science,2019,155:698-703.
[7]DARGAN S,KUMAR M,AYYAGARI M R,et al.A survey of deep learning and its applications:a new paradigm to machine learning[J].Archives of Computational Methods in Enginee-ring,2020,27(4):1071-1092.
[8]WANG Y,YAO Q,KWOK J T,et al.Generalizing from a few examples:A survey on few-shot learning[J].ACM Computing Surveys(csur),2020,53(3):1-34.
[9]WIJEKOON A,WIRATUNGA N.Personalised meta-learningfor human activity recognition with few-data[C]//International Conference on Innovative Techniques and Applications of Artificial Intelligence.Cham:Springer,2020:79-93.
[10]HUISMAN M,VAN RIJN J N,PLAAT A.A survey of deepmeta-learning[J].Artificial Intelligence Review,2021,54(6):4483-4541.
[11]FINN C,ABBEEL P,LEVINE S.Model-agnostic meta-learning for fast adaptation of deep networks[C]//International Confe-rence on Machine Learning.PMLR,2017:1126-1135.
[12]LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-basedlearning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278- 2324.
[13]WANG J,CHEN Y,HAO S,et al.Deep learning for sensor-based activity recognition:A survey[J].Pattern Recognition Letters,2019,119:3-11.
[14]BALDOMINOS A,CERVANTES A,SAEZ Y,et al.A comparison of machine learning and deep learning techniques for activity recognition using mobile devices[J].Sensors,2019,19(3):521.
[15]ZHAO Y,YANG R,CHEVALIER G,et al.Deep residual bidir-LSTM for human activity recognition using wearable sensors[J].Mathematical Problems in Engineering,2018,2018:7316954.
[16]MUTEGEKI R,HAN D S.A CNN-LSTM approach to human activity recognition[C]//2020 International Conference on Artificial Intelligence in Information and Communication(ICAIIC).IEEE,2020:362-366.
[17]ORDÓÑEZ F J,ROGGEN D.Deep convolutional and LSTM recurrent neural networks for multimodal wearable activity recognition[J].Sensors,2016,16(1):115.
[18]HAMMERLA N Y,HALLORAN S,PLÖTZ T.Deep,convolutional,and recurrent models for human activity recognition using wearables[J].arXiv:1604.08880,2016.
[19]NAFEA O,ABDUL W,MUHAMMAD G,et al.Sensor-based human activity recognition with spatio-temporal deep learning[J].Sensors,2021,21(6):2141.
[20]KHAN Z N,AHMAD J.Attention induced multi-head convolutional neural network for human activity recognition[J].Applied Soft Computing,2021,110:107671.
[21]MUHAMMAD K,ULLAH A,IMRAN A S,et al.Human action recognition using attention based LSTM network with dilated CNN features[J].Future Generation ComputerSystems,2021,125:820-830.
[22]THRUN S,LORIEN P,et al.Learning to learn[M].Springer Science & Business Media,2012.
[23]LAKE B M,SALAKHUTDINOV R,TENENBAUM J B.Human-level concept learning through probabilistic program induction[J].Science,2015,350(6266):1332-1338.
[24]RAVI S,LAROCHELLE H.Optimization as a model for few-shot learning [C]//International Conference on Learning Representations.2017.
[25]YAO H,LIU Y,WEI Y,et al.Learning from multi-ple cities:A meta-learning approach for spatialte-mporal prediction[C]//The World Wide Web Conference.2019:2181-2191.
[26]FENG S,DUARTE M F.Few-shot learning-based human acti-vity recognition[J].Expert Systems with Applications,2019,138:112782.
[27]SUNG F,YANG Y,ZHANG L,et al.Learning to compare:Relation network for few-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:1199-1208.
[28]WIJEKOON A,NIRMALIE W,KAY C.Mex:Multi-modalexercises dataset for human activity recognition[J].arXiv:1908.08992,2019.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发