计算机科学

计算机科学 ›› 2019, Vol. 46 ›› Issue (12): 306-312.doi: 10.11896/jsjkx.191200500C

• 交叉与前沿 • 上一篇    下一篇

基于多模融合的半监督场景识别方法

沈鸿1,2,3, 刘军发1,2,3, 陈益强1,2,3, 蒋鑫龙1,2,3, 黄正宇2,3   

  1. (中国科学院大学 北京100190)1;
    (中国科学院计算技术研究所泛在计算系统研究中心 北京100190)2;
    (北京市移动计算与新型终端重点实验室 北京100190)3
  • 收稿日期:2019-01-11 出版日期:2019-12-15 发布日期:2019-12-17
  • 通讯作者: 刘军发(1973-),男,博士,副研究员,CCF高级会员,主要研究方向为无线定位、机器学习,E-mail:liujunfa@ict.ac.cn。
  • 作者简介:沈鸿(1995-),男,硕士生,主要研究方向为普适计算、无线定位,E-mail:shenhong@ict.ac.cn;陈益强(1973-),男,博士,研究员,博士生导师,主要研究方向为普适计算、人机交互;蒋鑫龙(1989-),男,博士,助理研究员,主要研究方向为无线定位、机器学习;黄正宇(1991-),男,硕士,主要研究方向为无线定位、机器学习。
  • 基金资助:
    本文受国家自然科学基金面上项目(61572004,61472399,61572471)资助。

Semi-supervised Scene Recognition Method Based on Multi-mode Fusion

SHEN Hong1,2,3, LIU Jun-fa1,2,3, CHEN Yi-qiang1,2,3, JIANG Xin-long1,2,3, HUANG Zheng-yu2,3   

  1. (University of Chinese Academy of Sciences,Beijing 100190,China)1;
    (Research Center for Ubiquitous Computing Systems,Institute of Computing Technology,Chinese Academyof Sciences,Beijing 100190,China)2;
    (Beijing Key Laboratory of Mobile Computing and Pervasive Device,Chinese Academy of Sciences,Beijing 100190,China)3
  • Received:2019-01-11 Online:2019-12-15 Published:2019-12-17

PDF (PC)

1109

摘要： 场景识别是普适计算中的一项重要研究内容,旨在通过识别智能手机用户所在位置的场景,为用户提供精准的个性化服务并提升服务的质量。在实际环境中,精确的场景识别存在两个问题:(1)基于单模传感器数据或无线信号数据的分类效果不佳、普适性不足;(2)场景识别的精度需要依赖大量标定数据,导致成本较高。针对这些问题,提出一种基于多模融合的半监督场景识别方法,该方法充分利用Wi-Fi、蓝牙和传感器的多模特征来提高识别精度。相比基于单模数据的识别,融合特征将静态场景的分类精度提升了10%,并且本文通过构建半监督的学习方法解决了动态场景中数据采集成本高的问题,在将标定数据量减少一半的基础上将识别精度提高至90%以上。实验数据表明,在利用Wi-Fi、蓝牙、传感器的互补优势的基础上,引入半监督的学习方法能够提升场景识别的精确度且降低在某些场景下采集数据的成本,从而有效地提升了场景识别的精度和普适性。

关键词: Wi-Fi, 半监督学习, 场景识别, 多模融合, 蓝牙

Abstract: Scene recognition is an important part of pervasive computing.It aims to provide users with accurate persona-lized service and improve service quality by identifying the location of the smartphone users.In the actual environment,there are two problems in accurate scene recognition.Firstly,based on single mode sensor data or wireless signal data,classification effect is not good enough ,and its generalization is not enough.Secondly,scene recognition accuracy depends on a large number of labeled data,resulting in high cost.In view of these problems,a semi-supervised scene recognition method based on multi-mode fusion was proposed.The menthod makes full use of the complementary information of Wi-Fi,Bluetooth and sensors to improve the accuracy of recognition.Compared with the recognition based on single mode data,fusion feature can increase the static scene classification accuracy by 10%.In this paper,a semi-supervised learning method was constructed to solve the problem of high data acquisition cost in dynamic scene,and the classification accuracy is over 90% by reducing half of the labeled data.The results show that introducing semi-supervised lear-ning method based on the complementary advantages of Wi-Fi,Bluetooth and sensors information can reduce data collecting cost and improve scene recognition accuracy to some extent,and thus highly increase its recognition accuracy and universality.

Key words: Bluetooth, Multi-mode fusion, Scene recognition, Semi-supervised learning, Wi-Fi

中图分类号: 

  • TP311
[1]KAPITSAKI G M,PREZERAKOS G N,TSELIKAS N D, et al.Context-aware service engineering:A survey[J].Journal of Systems & Software,2009,82(8):1285-1297.
[2]CHEN Z Y.Mining individual behavior pattern based on significant locations and spatial trajectories[C]//IEEE International Conference on Pervasive Computing and Communications Workshops.Lugano:IEEE,2012:540-541.
[3]CHEN Z,ZHOU J,CHEN Y,et al.Combing multiple linear regression and manifold regularization for indoor positioning from unique radio signal[C]//Pervasive Computing.Taibei:IEEE,2009:611-614.
[4]CHEN Y Q,CHEN Z Y,LIU J F,et al.Surrounding context and episode awareness using dynamic Bluetooth data[C]//ACM Conference on Ubiquitous Computing.For Worth:ACM,2012:629-630.
[5]YE S Y,ZHANG W Z,QI T L,et al.A Sensor and User Beha- vior Data Analysis Based Method of Mobile Learning Situation Perception[J].Journal of Computer Research and Development,2016,52(12):2721-2728.(in Chinese)
叶舒雁,张未展,齐天亮,等.一种基于传感器与用户行为数据分析的移动学习场景感知分类方法[J].计算机研究与发展,2016,52(12):2721-2728.
[6]RUAN J J,LUO D,LUO H Y.Indoor and outdoor scene recognition algorithm based on support vector machine multi-classifier[J].Journal of Computer Application,2015,35(11):3135-3138.(in Chinese)
阮锦佳,罗丹,罗海勇.基于支持向量机多分类器的室内外场景感知算法[J].计算机应用,2015,35(11):3135-3138.
[7]YOUSSEF M A,AGRAWALA A,SHANKAR A U.WLAN Location Determination via Clustering and Probability Distributions[C]//IEEE International Conference on Pervasive Computing and Communications.IEEE,2003:143-150.
[8]KRUMM J,HARIHARAN R.TempIO:Inside/Outside Classification with Temperature[C]//Second International Workshop on Man-Machine Symbiotic Systems.Kyoto,2004.
[9]BI S C,YUAN K,CHEN J L,et al.A method of user’s scene recognition based on WIFI portrait:CN 201610373922.6.2016.(in Chinese)
毕树超,袁凯,陈津来,等.一种基于WIFI画像识别用户所在场景的方法:CN 201610373922.6.2016
[10]LIU Z W,XU F Y,WANG Z X.A new multi-user indoor location algorithm based on difference distance-loss model with parameter fitting[J].Chinese Journal of Radio Science,2008,23(6):1090-1094.(in Chinese)
刘召伟,徐凤燕,王宗欣.基于参数拟合的室内多用户定位算法[J].电波科学学报,2008,23(6):1090-1094
[11]MILETTE G,STROUD A.Professional Android Sensor Programming[M].Wrox Press Ltd,2012.
[12]ZHU X M,LI S R,LI Z B.WKNN Indoor Positioning Method Based on Distribution Overlap of Rssi[J].Journal of Xuzhou Institude of Technology(Natural Sciences Edition),2017,32(3):48-52.(in Chinese)
朱雪梅,李石荣,李泽彬.基于RSSI分布重叠的WKNN室内定位方法[J].徐州工程学院学报(自然科学版),2017,32(3):48-52
[13]HUANG G B,ZHU Q Y,SIEW C K.Extreme learning ma- chine:a new learning scheme of feedforward neural networks[C]//IEEE International Joint Conference on Neural Networks.Budapest:IEEE,2005:985-990.
[14]BANERJEE K S.Generalized Inverse of Matrices and Its Applications[J].Technometrics,1971,15(1):197-197.
[15]CVETKOVIC D M,DOOB M,SACHS H.Spectra of graphs:theory and application[J].Math Hungar,1995,6(2):191-195.
[16]CHERKASSKY V.The Nature Of Statistical Learning Theory [J].IEEE Transactions on Neural Networks,1997,8(6):1564-1564.
[17]JERIBI A.Spectral Graph Theory[M]//Spectral graph theory.Springer,1997:212.
[18]XU R,JIANG F,YAO H X.Overview of manifold learning [J].Journal of Intelligent Systems,2006,1(1):44-51.(in Chinese)
徐蓉,姜峰,姚鸿勋.流形学习概述[J].智能系统学报,2006,1(1):44-51.
[19]BELKIN M,MATVEEVA I,NIYOGI P.Regularization and Semi-supervised Learning on Large Graphs[C]//International Conference on Computational Learning Theory.Berlin:Sprin-ger,2004.
[20]BELKIN M,NIYOGI P,SINDHWANI V.Manifold Regularization:A Geometric Framework for Learning from Labeled and Unlabeled Examples[J].Journal of Machine Learning Research,2006,7(1):2399-2434.
[1] 武红鑫, 韩萌, 陈志强, 张喜龙, 李慕航.
监督和半监督学习下的多标签分类综述
Survey of Multi-label Classification Based on Supervised and Semi-supervised Learning
计算机科学, 2022, 49(8): 12-25. https://doi.org/10.11896/jsjkx.210700111
[2] 侯夏晔, 陈海燕, 张兵, 袁立罡, 贾亦真.
一种基于支持向量机的主动度量学习算法
Active Metric Learning Based on Support Vector Machines
计算机科学, 2022, 49(6A): 113-118. https://doi.org/10.11896/jsjkx.210500034
[3] 庞兴龙, 朱国胜.
基于半监督学习的网络流量分析研究
Survey of Network Traffic Analysis Based on Semi Supervised Learning
计算机科学, 2022, 49(6A): 544-554. https://doi.org/10.11896/jsjkx.210600131
[4] 王宇飞, 陈文.
基于DECORATE集成学习与置信度评估的Tri-training算法
Tri-training Algorithm Based on DECORATE Ensemble Learning and Credibility Assessment
计算机科学, 2022, 49(6): 127-133. https://doi.org/10.11896/jsjkx.211100043
[5] 许华杰, 秦远卓, 杨洋.
基于多级特征融合与注意力模块的场景识别方法
Scene Recognition Method Based on Multi-level Feature Fusion and Attention Module
计算机科学, 2022, 49(4): 209-214. https://doi.org/10.11896/jsjkx.210100135
[6] 许华杰, 陈育, 杨洋, 秦远卓.
基于混合样本自动数据增强技术的半监督学习方法
Semi-supervised Learning Method Based on Automated Mixed Sample Data Augmentation Techniques
计算机科学, 2022, 49(3): 288-293. https://doi.org/10.11896/jsjkx.210100156
[7] 刘昕, 袁家斌, 王天星.
基于场景先验知识的室内人体行为识别方法
Interior Human Action Recognition Method Based on Prior Knowledge of Scene
计算机科学, 2022, 49(1): 225-232. https://doi.org/10.11896/jsjkx.201100185
[8] 孔金生, 李婧馨, 段鹏松, 曹仰杰.
基于Wi-Fi信号的人体身份识别算法研究综述
Survey of Human Identification Algorithms Based on Wi-Fi Signal
计算机科学, 2021, 48(10): 246-257. https://doi.org/10.11896/jsjkx.210100076
[9] 秦悦, 丁世飞.
半监督聚类综述
Survey of Semi-supervised Clustering
计算机科学, 2019, 46(9): 15-21. https://doi.org/10.11896/j.issn.1002-137X.2019.09.002
[10] 吴振宇, 李云雷, 吴凡.
基于Tucker分解的半监督支持张量机
Semi-supervised Support Tensor Based on Tucker Decomposition
计算机科学, 2019, 46(9): 195-200. https://doi.org/10.11896/j.issn.1002-137X.2019.09.028
[11] 江泽涛, 秦嘉奇, 胡硕.
基于多路卷积神经网络的多光谱场景识别方法
Multi-spectral Scene Recognition Method Based on Multi-way Convolution Neural Network
计算机科学, 2019, 46(9): 265-270. https://doi.org/10.11896/j.issn.1002-137X.2019.09.040
[12] 李森森, 黄一才, 郁滨.
一种从设备零秘密存储的蓝牙密钥协商方案
Bluetooth Key Agreement Scheme with Zero Secret Storage in Slave Device
计算机科学, 2019, 46(4): 151-157. https://doi.org/10.11896/j.issn.1002-137X.2019.04.024
[13] 于怡然, 常俊, 吴柳繁, 张永鸿.
基于Wi-Fi信号的免训练呼吸检测
Training-free Human Respiration Sensing Based on Wi-Fi Signal
计算机科学, 2019, 46(11): 304-308. https://doi.org/10.11896/jsjkx.190600143
[14] 喻影, 陈珂, 寿黎但, 陈刚, 吴晓凡.
基于关键词和关键句抽取的用户评论情感分析
Sentiment Analysis of User Comments Based on Extraction of Key Words and Key Sentences
计算机科学, 2019, 46(10): 19-26. https://doi.org/10.11896/jsjkx.191000531C
[15] 刘枭, 王晓国.
基于概率图的银行电信诈骗检测方法
Probabilistic Graphical Model Based Approach for Bank Telecommunication Fraud Detection
计算机科学, 2018, 45(7): 122-128. https://doi.org/10.11896／j.issn.1002-137X.2018.07.020
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