TY - GEN
T1 - Spatial Gaussian mixture model for gender recognition
AU - Li, Zhen
AU - Zhou, Xi
AU - Huang, Thomas S
PY - 2009
Y1 - 2009
N2 - Patch-based approaches have become popular in many computer vision applications over recent years. An intrinsic flaw of this framework, missing of the spatial information, however, restricts its usage in face related applications where the spatial configuration is relatively settled. In this paper, we introduce a new patch feature representation, namely Spatial Gaussian Mixture Models (SGMM), which enhances the traditional GMM approach by taking the spatial information into consideration at both local and global scales. In the meantime, SGMM inherits all the merits of GMM, such as precise appearance description and robustness to image mis-alignment. The experiments on gender recognition demonstrate that the SGMM representation achieves more than 40% relative error reduction compared with either GMM or SVM-based approaches.
AB - Patch-based approaches have become popular in many computer vision applications over recent years. An intrinsic flaw of this framework, missing of the spatial information, however, restricts its usage in face related applications where the spatial configuration is relatively settled. In this paper, we introduce a new patch feature representation, namely Spatial Gaussian Mixture Models (SGMM), which enhances the traditional GMM approach by taking the spatial information into consideration at both local and global scales. In the meantime, SGMM inherits all the merits of GMM, such as precise appearance description and robustness to image mis-alignment. The experiments on gender recognition demonstrate that the SGMM representation achieves more than 40% relative error reduction compared with either GMM or SVM-based approaches.
KW - KL-divergence
KW - SGMM
KW - UBM
UR - http://www.scopus.com/inward/record.url?scp=77951940789&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=77951940789&partnerID=8YFLogxK
U2 - 10.1109/ICIP.2009.5413917
DO - 10.1109/ICIP.2009.5413917
M3 - Conference contribution
AN - SCOPUS:77951940789
SN - 9781424456543
T3 - Proceedings - International Conference on Image Processing, ICIP
SP - 45
EP - 48
BT - 2009 IEEE International Conference on Image Processing, ICIP 2009 - Proceedings
PB - IEEE Computer Society
T2 - 2009 IEEE International Conference on Image Processing, ICIP 2009
Y2 - 7 November 2009 through 10 November 2009
ER -