[ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ]

Journal of Information Science and Engineering, Vol. 34 No. 5, pp. 1237-1249

Robust Facial Expression Recognition using Local Haar Mean Binary Pattern

1,2Department of Computer Engineering
1Charotar University of Science and Technology
Changa, 388421 India

2Gujarat Technological University
V. V. Nagar, 388120 India
E-mail: mgoyani@gmail.com, nmpatel@bvmengineerring.ac.in

  In this paper, we propose a hybrid statistical feature extractor, Local Haar Mean Binary Pattern (LHMBP). It extracts level-1 haar approximation coefficients and computes Local Mean Binary Pattern (LMBP) of it. LMBP code of pixel is obtained by weighting the thresholded neighbor value of 3×3 patch on its mean. LHMBP produces highly discriminative code compared to other state of the art methods. To localize appearance features, approximation subband is divided into M×N regions. LHMBP feature descriptor is derived by concatenating LMBP distribution of each region. We also propose a novel template matching strategy called Histogram Normalized Absolute Difference (HNAD) for histogram based feature comparison. Experiments prove the superiority of HNAD over well-known template matching techniques such as L2 norm and Chi-Square. We also investigated LHMBP for expression recognition in low resolution. The performance of the proposed approach is tested on well-known CK, JAFFE, and SFEW facial expression datasets in diverse situations.

Keywords: affective computing, appearance based feature, local binary pattern, Gabor filter, support vector machine

  Retrieve PDF document (JISE_201805_08.pdf)