11th EAI International Conference on Mobile Multimedia Communications

Research Article

The Empirical Study On Deep Convolutional Network Transferring Among Users Within Activity Recognition

Download629 downloads
  • @INPROCEEDINGS{10.4108/eai.21-6-2018.2276633,
        author={Jiazhen Li and Renjie Ding and Lanshun Nie and Xue Li and Xiandong Si and Lulu Wang and Dianhui Chu and Dechen Zhan},
        title={The Empirical Study On Deep Convolutional Network Transferring Among Users Within Activity Recognition},
        proceedings={11th EAI International Conference on Mobile Multimedia Communications},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2018},
        month={9},
        keywords={sensor-based data human activity recognition cnn transfer learning},
        doi={10.4108/eai.21-6-2018.2276633}
    }
    
  • Jiazhen Li
    Renjie Ding
    Lanshun Nie
    Xue Li
    Xiandong Si
    Lulu Wang
    Dianhui Chu
    Dechen Zhan
    Year: 2018
    The Empirical Study On Deep Convolutional Network Transferring Among Users Within Activity Recognition
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.21-6-2018.2276633
Jiazhen Li1, Renjie Ding1, Lanshun Nie1,*, Xue Li1, Xiandong Si1, Lulu Wang2, Dianhui Chu1, Dechen Zhan1
  • 1: Harbin Institute of Technology
  • 2: Dalian University
*Contact email: nls@hit.edu.cn

Abstract

Human activity recognition based on sensor data is one of the significant problems in pervasive computing. In recent years, deep learning has become the main method in this field due to its high accuracy. However, it’s difficult to recognize activities of user B with the model trained for user A. The effect of transferring the model (among different users) is the key that restricts activity recognition in practice. At present, there is still little research on the transferring of deep learning model in this field. Its effect, principle and influencial factors remain to be studied. We carried out the empirical study on the transferring of deep learning model among users. We visualized the features extracted from CNN and studied its distribution. We compared the feasibility, strength and weakness of typical unsupervised and semi-supervised transferring methods. The observations and insights in this study have deepened the understanding of transferring in activity recognition field and provide guidance for further research.