Laser & Optoelectronics Progress, Volume. 57, Issue 10, 101508(2020)
A Convolutional Neural Network Based on Feature Fusion for Face Recognition
Figures & Tables(10)
Fig. 2. DCT transform and reconstruction of face image. (a) Original image (128×128); (b) DCT energy distribution map; (c) face image reconstructed with coefficients 24×24; (d) face image reconstructed with coefficients 48×48
Table 1. Recognition rate based on ORL face database
View tableTable 1. Recognition rate based on ORL face database
DCT coefficient a1=0.9, a2=0.1 a1=0.85, a2=0.15 a1=0.8, a2=0.2 a1=0.7, a2=0.3 a1=0.6, a2=0.4 a1=0.5, a2=0.5 8×8 16×16 24×24 32×32 40×40 48×48 56×56 64×64 0.93125 0.95938 0.96250 0.98125 0.96875 0.97188 0.92187 0.92500 0.95000 0.94688 0.96250 0.98438 0.97500 0.94375 0.95625 0.96875 0.91563 0.94687 0.94375 0.96875 0.95313 0.93438 0.96563 0.91562 0.93125 0.96875 0.97188 0.97500 0.95313 0.97812 0.93750 0.90625 0.96875 0.96250 0.93125 0.97187 0.95625 0.95312 0.94376 0.95938 0.94375 0.92187 0.94063 0.96817 0.95469 0.95938 0.95000 0.95469
Table 2. Recognition rate based on CAS-PEAL face database
View tableTable 2. Recognition rate based on CAS-PEAL face database
DCT coefficient a1=0.9, a2=0.1 a1=0.85, a2=0.15 a1=0.8, a2=0.2 a1=0.7, a2=0.3 a1=0.6, a2=0.4 a1=0.5, a2=0.5 8×8 16×16 24×24 32×32 40×40 48×48 56×56 64×64 0.90365 0.94358 0.96181 0.96875 0.93229 0.94531 0.96701 0.96747 0.97625 0.98292 0.97438 0.98312 0.97562 0.96313 0.97021 0.96750 0.96094 0.92361 0.90972 0.97541 0.94542 0.97021 0.96441 0.94097 0.95771 0.94705 0.94917 0.97125 0.95792 0.96771 0.95229 0.96875 0.96938 0.97354 0.95486 0.97354 0.96625 0.94358 0.95920 0.92792 0.95573 0.95521 0.93604 0.96958 0.95658 0.96687 0.95418 0.94531
Table 3. Recognition rate comparison of four algorithms
View tableTable 3. Recognition rate comparison of four algorithms
Database Algorithm CNN DCT-CNN LBP-CNN LBP-DCT-CNN ORL CAS-PEAL 0.86563 0.92333 0.95313 0.97687 0.91250 0.93750 0.98438 0.98312
Tools
Get Citation
Copy Citation Text
Jiaxin Wang, Zhichun Lei. A Convolutional Neural Network Based on Feature Fusion for Face Recognition[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101508
Paper Information
Category: Machine Vision
Received: Aug. 29, 2019
Accepted: Oct. 22, 2019
Published Online: May. 8, 2020
The Author Email: Jiaxin Wang (18712768825@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20