Laser & Optoelectronics Progress, Volume. 57, Issue 8, 081009(2020)

Identity Authentication for Smart Phones Based on an Optimized Convolutional Deep Belief Network

Yichao Zhang1 and Ziwen Sun1,2、*
Author Affiliations
  • 1School of Internet of Things, Jiangnan University, Wuxi, Jiangsu 214122, China
  • 2Engineering Research Center of Internet of Things Technology Applications of Ministry of Education, Wuxi, Jiangsu 214122, China
  • show less
    Figures & Tables(15)
    Overall framework of gesture identity authentication
    Different types of CDBN models structure. (a) General structure; (b) with pooling layer
    Different model structures. (a) RBM structure; (b) CRBM structure
    Maximum pooling operation
    Single step Gibbs sampling
    Structure diagram of the output authentication model
    User 1 original motion trajectory
    User 1 pre-processed trajectory
    Fake user's gesture recovery diagram
    • Table 1. Simulation results of different network depths

      View table

      Table 1. Simulation results of different network depths

      DepthACC /%FAR /%FRR /%Time /s
      193.5246.228.0060.90
      297.6672.223.33130.56
      392.3337.1110.67251.12
    • Table 2. Simulation results of different sparsity indices

      View table

      Table 2. Simulation results of different sparsity indices

      IndexACC /%FAR /%FRR /%Time /s
      0.01096.3332.808.00109.36
      0.01597.0002.008.00115.98
      0.02097.6672.223.33130.56
      0.02595.000030.00134.02
      0.20083.3330100.00139.88
    • Table 3. Simulation results of different pooling methods

      View table

      Table 3. Simulation results of different pooling methods

      MethodACC /%FAR /%FRR /%Time /s
      None97.3332.006.00400.31
      Mean97.6522.333.35130.77
      Max97.6672.223.33130.56
    • Table 4. Simulation results of different iteration times

      View table

      Table 4. Simulation results of different iteration times

      EpochACC /%FAR /%FRR /%Time /s
      596.0000.8020.0070.33
      1097.6672.223.33130.56
      2596.0002.8010.00375.61
      5097.0002.008.00643.80
    • Table 5. Simulation results of different connection layers

      View table

      Table 5. Simulation results of different connection layers

      LayerACC /%FAR /%FRR /%Time /s
      Fully97.3332.672.67367.44
      RMS97.6672.223.33130.56
    • Table 6. Performance comparison among CDBN algorithm, BP, HMM, and DBN algorithms

      View table

      Table 6. Performance comparison among CDBN algorithm, BP, HMM, and DBN algorithms

      MethodACC /%FAR /%FRR /%Time /s
      BP92.6604.535.0156.84
      HMM93.2503.444.4685.69
      DBN96.6302.373.79119.31
      CDBN97.6672.223.33130.56
    Tools

    Get Citation

    Copy Citation Text

    Yichao Zhang, Ziwen Sun. Identity Authentication for Smart Phones Based on an Optimized Convolutional Deep Belief Network[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081009

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Jul. 23, 2019

    Accepted: Sep. 11, 2019

    Published Online: Apr. 3, 2020

    The Author Email: Ziwen Sun (sunziwen@jiangnan.edu.cn)

    DOI:10.3788/LOP57.081009

    Topics