ZTE

ZTE Communications ›› 2016, Vol. 14 ›› Issue (S0): 59-66.doi: 10.3969/j.issn.1673-5188.2016.S0.008

• Research Paper • Previous Articles    

Human Motion Recognition Based on Incremental Learning and Smartphone Sensors

LIU Chengxuan1, DONG Zhenjiang2, XIE Siyuan2, PEI Ling1   

  1. 1. Shanghai Jiao Tong University,Shanghai 200240,China;
    2. ZTE Corporation,Shenzhen 518057,China
  • Received:2015-07-20 Online:2016-06-01 Published:2019-11-29
  • About author:LIU Chengxuan (lcxstorm@163.com) is studying for the ME degree in the Department of Information and Communication Engineering, Shanghai Jiao Tong University, China. He received his BE degree from Shanghai Jiao Tong University. His research interests include pattern recognition, multisource navigation, and positioning technology.
    DONG Zhenjiang (dong.zhenjiang@zte.com.cn) is the vice president of the Cloud Computing & IT Research Institute of ZTE Corporation. His main research areas are cloud computing, big data, new media, and mobile internet technologies.
    XIE Siyuan (xie.siyuan7@zte.com.cn) is a pre-research engineer at ZTE Corporation. His main research areas are indoor positioning, IoT, and mobile internet technologies.
    PEI Ling (ling.pei@sjtu.edu.cn) is an associate professor in the Department of Electronic Engineering, Shanghai Jiao Tong University, China. He received his PhD degree from Southeast University, China. His research interests include indoor and outdoor seamless positioning, context-aware technology, and motion pattern recognition.
  • Supported by:
    This work is partly supported by the National Natural Science Foundation of China under Grant 61573242; the Projects from Science and Technology Commission of Shanghai Municipality under Grant No.13511501302,No.14511100300,and No.15511105100; Shanghai Pujiang Program under Grant No.14PJ1405000; ZTE Industry-Academia-Research Cooperation Funds

PDF (PC)

Abstract: Batch processing mode is widely used in the training process of human motion recognition. After training, the motion classifier usually remains invariable. However, if the classifier is to be expanded, all historical data must be gathered for retraining. This consumes a huge amount of storage space, and the new training process will be more complicated. In this paper, we use an incremental learning method to model the motion classifier. A weighted decision tree is proposed to help illustrate the process, and the probability sampling method is also used. The results show that with continuous learning, the motion classifier is more precise. The average classification precision for the weighted decision tree was 88.43% in a typical test. Incremental learning consumes much less time than the batch processing mode when the input training data comes continuously.

Key words: human motion recognition, incremental learning, mapping function, weighted decision tree, probability sampling