Laser & Optoelectronics Progress, Volume. 55, Issue 3, 031008(2018)

Real-Time Gesture Recognition Based on Kinect

Zhiqiang Bao* and Chengang Lü
Author Affiliations
  • School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
  • show less
    Figures & Tables(6)
    Processing of gesture region extraction. (a) Gesture image; (b) result of arm-removing
    Three circular LBP descriptors
    Ten kinds of Chinese sign language gestures
    Confusion matrix for features of (a) R-LBP, (b) U-LBP, and (c) HOG
    • Table 1. Time-consuming comparison of three recognition methods

      View table

      Table 1. Time-consuming comparison of three recognition methods

      MethodAverage time offeatureextraction /msTraining time /sTest time /sAveragerecognition time /msFeature+recognition /ms
      R-LBP+SVM11.25282.657.07.1318.68
      U-LBP+SVM14.5730.630.63.8318.20
      HOG+SVM10.354.84.20.5310.83
    • Table 2. Performance comparison of proposed method with benchmark methods

      View table

      Table 2. Performance comparison of proposed method with benchmark methods

      MethodSegmentation methodFeatureRecognitionaccuracy /%Averagerunning time /ms
      Method in Ref. [6]HybridHu+fingertips+trajectory97.5-
      Method in Ref. [7]Depth threshold+reference vectorEPS94.1689
      Method in Ref. [8]HybridFEMD93.275
      Method in Ref. [17]Depth+CL TreeShape+position+orientation90.35293
      Method in Ref. [18]Depth+KFCMHOG96-9748.1
      Method in Ref. [19]Depth+k-curvatureFingertips9133.3
      Proposed methodDepth+KalmanR-LBP78.5839.73
      U-LBP91.5839.25
      HOG97.0931.95
    Tools

    Get Citation

    Copy Citation Text

    Zhiqiang Bao, Chengang Lü. Real-Time Gesture Recognition Based on Kinect[J]. Laser & Optoelectronics Progress, 2018, 55(3): 031008

    Download Citation

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

    Category: Image processing

    Received: Sep. 5, 2017

    Accepted: --

    Published Online: Sep. 10, 2018

    The Author Email: Bao Zhiqiang (bzq1028@tju.edu.cn)

    DOI:10.3788/LOP55.031008

    Topics