TY - JOUR

T1 - Improving face recognition using deep autoencoders and feature fusion

AU - Khider, A.

AU - Djemili, R.

AU - Bouridane, A.

AU - Jiang, R.

PY - 2023/1/30

Y1 - 2023/1/30

N2 - Uncontrolled environments are the main challenges of real face recognition systems, recent success of deep learning and features fusion has led to various performance improvements. This paper proposes a novel scheme called feature autoencoder (FAE), where an autoencoder model is not trained directly from the raw facial images, rather it uses a fusion of features constructed by Gabor filter, local binary pattern and local phase quantisation. For each feature, a linear discriminant analysis is applied to reduce its high dimensionality and a limited adaptive histogram equalisation process is employed for contrast enhancement. The proposed scheme has been evaluated using known datasets such as AR, ORL and YALE, and the experimental results carried out on these databases have been compared using three classifiers: k-nearest neighbour, multiclass support vector machine and softmax classifier, demonstrating the effectiveness of proposed approach and parameters. The experimental results obtained and compared with recent and similar approaches on six databases: ORL, YALE, AR, extended YALE B, CMU PIE, and LFWcrop, suggest that the proposed technique outperforms similar techniques. The recognition rates got from them are 100%, 100%, 99.66%, 99.40%, 97.31%, and 90.68% respectively.

AB - Uncontrolled environments are the main challenges of real face recognition systems, recent success of deep learning and features fusion has led to various performance improvements. This paper proposes a novel scheme called feature autoencoder (FAE), where an autoencoder model is not trained directly from the raw facial images, rather it uses a fusion of features constructed by Gabor filter, local binary pattern and local phase quantisation. For each feature, a linear discriminant analysis is applied to reduce its high dimensionality and a limited adaptive histogram equalisation process is employed for contrast enhancement. The proposed scheme has been evaluated using known datasets such as AR, ORL and YALE, and the experimental results carried out on these databases have been compared using three classifiers: k-nearest neighbour, multiclass support vector machine and softmax classifier, demonstrating the effectiveness of proposed approach and parameters. The experimental results obtained and compared with recent and similar approaches on six databases: ORL, YALE, AR, extended YALE B, CMU PIE, and LFWcrop, suggest that the proposed technique outperforms similar techniques. The recognition rates got from them are 100%, 100%, 99.66%, 99.40%, 97.31%, and 90.68% respectively.

KW - autoencoder

KW - deep learning

KW - face recognition

KW - feature extraction

KW - fusion

KW - sparse

KW - uncontrolled environments

KW - Deep learning

KW - Discriminant analysis

KW - Gabor filters

KW - Learning systems

KW - Local binary pattern

KW - Nearest neighbor search

KW - Support vector machines

KW - Auto encoders

KW - Face recognition systems

KW - Facial images

KW - Features extraction

KW - Features fusions

KW - Fusion of features

KW - Performance

KW - Sparse

KW - Uncontroled environment

KW - Face recognition

U2 - 10.1504/IJBM.2022.10043147

DO - 10.1504/IJBM.2022.10043147

M3 - Journal article

VL - 15

SP - 40

EP - 58

JO - International Journal of Biometrics

JF - International Journal of Biometrics

IS - 1

ER -