Fig 1 - uploaded by Ayyad Maafiri
Content may be subject to copyright.
Source publication
Biometrics has become fashionable in areas that require a high level of security and control. Among all the technologies that exist,
face recognition is one of the most used and adapted technologies.In this work, we propose a new fusion of two projection based
face recognition algorithms in Discrete wavelet transform domain(DWT).Those two algorithm...
Contexts in source publication
Context 1
... main advantage of the wavelet transform over the Fourier transform is the location in time-scale. An image is decomposed into four subbands as shown in Figure 1. The band LL is a coarser approximation to the original image. ...
Similar publications
In this paper, we study the estimation of a rank-one spiked tensor in the presence of heavy tailed noise. Our results highlight some of the fundamental similarities and differences in the tradeoff between statistical and computational efficiencies under heavy tailed and Gaussian noise. In particular, we show that, for $p$ th order tensors, the trad...
Citations
... Almotiri, et al [12] proposed face Recognition using PCA and Clustered Self-Organizing Map. A new combination of two face recognition algorithms based on projection It is suggested by [13] to use singular value decomposition (SVD) and relevance weighted linear discriminant analysis (RW-LDA) using the left and right singular vectors. The application of a well-known moment, Legendre, to a vector of features generated by the singular value decomposition transform (SVD) proposed in [14] led to the fusion of features. ...
Face recognition is essential component of various applications including computer vision, security systems and biometrics. By examining the efficacy of several dimensionality reduction techniques, including Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Singular Value Decomposition (SVD), and Non-Negative Matrix Factorization (NMF), this paper offers a novel approach to face recognition. These techniques are combined with diverse classifiers, including Support Vector Machines (SVM), Random Forest, LightGBM, and k-Nearest Neighbors (KNN), are employed to evaluate their impact on face recognition accuracy. Experiments were conducted on Olivetti faces data set. We have demonstrated the comparative analysis of different dimensionality reduction techniques classifiers in terms of accuracy, precision, recall, f1score. Results shows the potential of integrating PCA with diverse classification models to enhance face recognition accuracy and highlights its applicability in real-world scenarios.
... Hence, constructing a discriminatory subspace that can differentiate between faces has received significant attention, and various strategies have been proposed, such as regularized LDA (RLDA) [5] and relevance-weighted LDA (RW-LDA) [6]. Each of these methods improves recognition accuracy and addresses the issue of small sample size (SSS) in different ways [1][2][3][4][5][6][7]. Another noteworthy feature extraction algorithm is Singular value decomposition (SVD), which is a matrix factorization mechanism that performs PCA. ...
... Another noteworthy feature extraction algorithm is Singular value decomposition (SVD), which is a matrix factorization mechanism that performs PCA. SVD has also impacted image processing and computer vision, particularly in the DWT domain, where it extracts reliable facial features in challenging lighting and occlusion conditions [7]. Maafiri et al. [7] proposed a new framework that fuses SVD with RW-LDA to achieve good recognition accuracy. ...
... SVD has also impacted image processing and computer vision, particularly in the DWT domain, where it extracts reliable facial features in challenging lighting and occlusion conditions [7]. Maafiri et al. [7] proposed a new framework that fuses SVD with RW-LDA to achieve good recognition accuracy. In this paper, we propose a new method for robust face recognition. ...
In this paper, we propose a new extension of the score level fusion of SVD (singular value decomposition) and RWLDA (relevance weighted linear discriminant analysis using QR decomposition) algorithms for face recognition called Improved DWT (SVD + RWLDA). The proposed extension exploits new techniques based on the symmetric sum using triangular norms. Experiments on two established data sets demonstrate that our approach enhances the face recognition rate compared to the original version.
... M. Ayyad et al. [15] developed a new discrete wavelet transform (DWT) domain fusion of two projection-based face recognition methods. These two algorithms use left and right singular vectors to perform linear fit weighted discriminant analysis (RW-LDA) and singular value decomposition (SVD). ...
... The general idea of hybrid systems is to combine face features generated by several feature extraction methods in order to minimize the effect of these variations as much as possible to have the best recognition rates. For exemple, the authors in [29] proposed a fusion method of two efficient dimensionnalityreduction techniques in discrete wavelet-transform domain. The application of this fusion with Min-Max as a method of normalization on two face datasets gives convincing results. ...
... The recognition rate of the LWKPCA method are compared with several recent advanced face recognition methods. Table 6 displays the recognition rate of our proposed approach compared with several recent face recognition methods including DIWTLBP [56], CFLDA [57], WTPCA-L1 [52], Gabor+SRC [58], LDA+SRC [58], DWT(SVD/LR+RWLDA/QR) [29], RKPCA [42], 2D-DMWT [59], IKLDA+PNN [60], DCT+VQ [61], Ga-bor+FastICA+LDA [31]. The proposed method adopts the same experimental-protocol of these methods. ...
... Table 7 displays the recognition rate of our proposed approach compared with several recent face recognition methods including ESGK [62] , Gabor filter (5×8)+MDC [62], LDA+MDC [62], NCDB-LBPc [9], DCD [10], Gabor+FastICA+LDA [31], Joint Collaborative [63], TPWELRMC [64]. CZZBP [8], CMBZZBP [8], CLBP [65], 6×6 MB-LBP [65], SAPFR-1 [66], INNC [67], SVD based VR [67], WTPCA-L1 [52], Naive CR [68], Method based on CR [68], RNLRLSR [69], CLSR [69], DWT(SVD/LR+RWLDA/QR) [29], DeepWTPCA-L 1 [30]. From this Table, we can notice that the proposed method LWKPCA achieved high recognition rates compared with the other methods. ...
Over the last two decades, face recognition (FR) has become one of the most prevailing biometric applications for effective people identification as it offers practical advantages over other biometric modalities. However, current state-of-the-art findings suggest that FR under adverse and challenging conditions still needs improvements. This is because face images can contain many variations like face expression, pose, and illumination. To overcome the effect of these challenges, it is necessary to use representative face features using feature extraction methods. In this paper, we present a new feature extraction method for robust FR called Local Binary Pattern and Wavelet Kernel PCA (LWKPCA). The proposed method aims to extract the discriminant and robust information to minimize recognition errors. This is obtained first by the best use of nonlinear projection algorithm called RKPCA. Then, we adapted the algorithm to reduce the dimensionality of features extracted using the proposed Color Local Binary Pattern and Wavelets transformation called Color LBP and Wavelet Descriptor. The general idea of our descriptor is to find the best representation of face image in a discriminant vector structure by a novel feature grouping strategy generated by the Three-Level decomposition of Discrete Wavelet Transform (2D-DWT) and Local Binary Pattern (LBP). Extensive experiments on four well-known face datasets namely ORL, GT, LFW , and YouTube Celebrities show that the proposed method has a recognition accuracy of 100% for ORL, 96.84% for GT, 99.34% for LFW, and 95.63% for YouTube Celebrities.
... We have also compared the performances of the proposed system with 9 state-of-the-art approaches on GTFD database and adopted the same experimental protocol. Table 4 shows the comparison of the recognition rate between our system proposed and these methods including SVD based VR [46], INNC [46], Naive CR [47], Method based on CR [47], RNL-RLSR [48], CLSR [48], DWT(SVD/LR+RWLDA/QR) using MIN-MAX method [49], DWT(SVD/LR+RWLDA/QR) using Z-score method [49], and CMBZZBP [50]. We can ...
... We have also compared the performances of the proposed system with 9 state-of-the-art approaches on GTFD database and adopted the same experimental protocol. Table 4 shows the comparison of the recognition rate between our system proposed and these methods including SVD based VR [46], INNC [46], Naive CR [47], Method based on CR [47], RNL-RLSR [48], CLSR [48], DWT(SVD/LR+RWLDA/QR) using MIN-MAX method [49], DWT(SVD/LR+RWLDA/QR) using Z-score method [49], and CMBZZBP [50]. We can ...
... Recognition Rate SVD based VR (2018) [46] 64.40% INNC (2018) [46] 63.60% Naive CR (2020) [47] 64.00% Method based on CR (2020) [47] 74.40% RNLRLSR (2020) [48] 72.00% CLSR (2020) [48] 65.60% DWT(SVD/LR+RWLDA/QR)+MIN-MAX (2019) [49] 89.24% DWT(SVD/LR+RWLDA/QR)+Z-score (2019) [49] 88.20% CMBZZBP (2020) [50] 91.20% DeepWTPCA-L 1 93.89% ...
In this paper, we propose a robust face recognition model called DeepWTPCA-L1 using WTPCA-L1 features and a CNN-LSTM architecture. First, WTPCA-L1 algorithm, composed of Three-level
decomposition of discrete wavelet transform followed by PCA-L1 algorithm, is exploited to extract face features. Then, the extracted features are used as inputs of the proposed CNN-LSTM architecture. To evaluate the robustness of the proposed approach, several face recognition datasets have been used. In addition, the proposed method is trained on noisy images using Gaussian, and Salt & Pepper noise added to the facial images of each dataset. The results of the experiment indicate that the proposed model achieves high recognition performance on three well-known standard face databases. When compared to state-of-the-art methods, the proposed model achieves a better face recognition rate.
... Typically, object recognition algorithms operate in two stages [3,54]. The first stage extracts features from the image using methods such as LBP [55][56][57], deformable part-based models (DPM) [58,59], or a histogram of oriented gradients (HOG) [60][61][62]. The second stage uses the feature vector to label the image, using classification methods such as nearest neighbors [63,64], support vector machines (SVM) [64,65], or deep neural networks [66,67]. ...
In this paper, we present the architecture of a smart imaging sensor (SIS) for face recognition, based on a custom-design smart pixel capable of computing local spatial gradients in the analog domain, and a digital coprocessor that performs image classification. The SIS uses spatial gradients to compute a lightweight version of local binary patterns (LBP), which we term ringed LBP (RLBP). Our face recognition method, which is based on Ahonen’s algorithm, operates in three stages: (1) it extracts local image features using RLBP, (2) it computes a feature vector using RLBP histograms, (3) it projects the vector onto a subspace that maximizes class separation and classifies the image using a nearest neighbor criterion. We designed the smart pixel using the TSMC 0.35 μm mixed-signal CMOS process, and evaluated its performance using postlayout parasitic extraction. We also designed and implemented the digital coprocessor on a Xilinx XC7Z020 field-programmable gate array. The smart pixel achieves a fill factor of 34% on the 0.35 μm process and 76% on a 0.18 μm process with 32 μm × 32 μm pixels. The pixel array operates at up to 556 frames per second. The digital coprocessor achieves 96.5% classification accuracy on a database of infrared face images, can classify a 150×80-pixel image in 94 μs, and consumes 71 mW of power.
... Ayyad et al. [21] presented a hybrid approach to face recognition based on the fusion of Singular Value Decomposition (SVD) and Relevance Weighted Linear Discriminant Analysis (RW-LDA) in Discrete wavelet transform domain. Their fusion method was employed to extract the significant features from facial images. ...
... It is evident from the results in Tab. 2 that the proposed method has obtained the highest recognition rate compared to the approaches presented in [13][14][15][16][17][20][21][22] across ORL face dataset. For instance, when the number of training samples is five, our proposed method provides a recognition rate of 99.10% which is significantly higher than those provided by other methods: 97.00% by [15], 94.50% and 97.85% by [17], 97.75% by [21], and 97.50% and 95.50% by [22]. ...
... It is evident from the results in Tab. 2 that the proposed method has obtained the highest recognition rate compared to the approaches presented in [13][14][15][16][17][20][21][22] across ORL face dataset. For instance, when the number of training samples is five, our proposed method provides a recognition rate of 99.10% which is significantly higher than those provided by other methods: 97.00% by [15], 94.50% and 97.85% by [17], 97.75% by [21], and 97.50% and 95.50% by [22]. The improvement noticed in the recognition rate by our new method is about 1.25, 2.19, and 1.13 points compared to the highest recognition rate provided by other methods across the three N-training samples (N = 5, 6, 8), respectively. ...
Over the past few decades, face recognition has become the most effective biometric technique in recognizing people's identity, as it is widely used in many areas of our daily lives. However, it is a challenging technique since facial images vary in rotations, expressions, and illuminations. To minimize the impact of these challenges, exploiting information from various feature extraction methods is recommended since one of the most critical tasks in face recognition system is the extraction of facial features. Therefore, this paper presents a new approach to face recognition based on the fusion of Gabor-based feature extraction, Fast Independent Component Analysis (Fas-tICA), and Linear Discriminant Analysis (LDA). In the presented method, first, face images are transformed to grayscale and resized to have a uniform size. After that, facial features are extracted from the aligned face image using Gabor, FastICA, and LDA methods. Finally, the nearest distance classifier is utilized to recognize the identity of the individuals. Here, the performance of six distance classifiers, namely Euclidean, Cosine, Bray-Curtis, Mahalanobis, Correlation, and Manhattan, are investigated. Experimental results revealed that the presented method attains a higher rank-one recognition rate compared to the recent approaches in the literature on four benchmarked face datasets: ORL, GT, FEI, and Yale. Moreover, it showed that the proposed method not only helps in better extracting the features but also in improving the overall efficiency of the facial recognition system.
... As a data processing method, SVD has been successfully applied to signal denoising processing and proved to be effective in avoiding modal aliasing. The main application of singular value decomposition in statistics is the principal component analysis (PCA), which can be used in pattern recognition [8][9][10][11], data dimensionality reduction [12,13], filter design [14,15], denoising [16,17], feature extraction and weak signal separation [18][19][20][21][22], face recognition [23,24] and many other fields, and it has gained important applications. In10, the application of SVD in pattern recognition of partial discharge signals is studied. ...
Reducing noise pollution in signals is of great significance in the field of signal detection. In order to reduce the noise in the signal and improve the signal-to-noise ratio (SNR), this paper takes the singular value decomposition theory as the starting point, and constructs various singular value decomposition denoising models with multiple multi-division structures based on the two-division recursion singular value decomposition, and conducts a noise reduction analysis on two experimental signals containing noise of different power. Finally, the SNR and mean square error (MSE) are used as indicators to evaluate the noise reduction effect, it is verified that the two-division recursion singular value decomposition is the optimal noise reduction model. This noise reduction model is then applied to the diagnosis of faulty bearings. By this method, the fault signal is decomposed to reduce noise and the detail signal with maximum kurtosis is extracted for envelope spectrum analysis. Comparison of several traditional signal processing methods such as empirical modal decomposition (EMD), ensemble empirical mode decomposition (EEMD), variational mode decomposition (VMD), wavelet decomposition, etc. The results show that multi-resolution singular value decomposition (MRSVD) has better noise reduction effect and can effectively diagnose faulty bearings. This method is promising and has a good application prospect.
... This study employed two feature reduction frameworks (LDA and Canonical Correlation Analysis [CCA]) and seven common classifiers. Ayyad and Khalid [16] suggested a novel combination of two projection-based FR frameworks in the DWT domain. The hybridized algorithms are RW-LDA and SVD using the left and right singular vectors. ...
... LDA is an important linear feature extraction method [16,23] that aims to find a set of the most discriminant projective vectors for mapping HD samples onto an LD space. All of the samples in the projective feature space will have the maximum inter-class scatter and minimum intra-class scatter, while the remaining test samples from the other classes will be easy to classify. ...
The big data concept has elicited studies on how to accurately and efficiently extract valuable information from such huge dataset. The major problem during big data mining is data dimensionality due to a large number of dimensions in such datasets. This major consequence of high data dimensionality is that it affects the accuracy of machine learning (ML) classifiers; it also results in time wastage due to the presence of several redundant features in the dataset. This problem can be possibly solved using a fast feature reduction method. Hence, this study presents a fast HP-PL which is a new hybrid parallel feature reduction framework that utilizes spark to facilitate feature reduction on shared/distributed-memory clusters. The evaluation of the proposed HP-PL on KDD99 dataset showed the algorithm to be significantly faster than the conventional feature reduction techniques. The proposed technique required 1 minute to select 4 dataset features from over 79 features and 3,000,000 samples on a 3-node cluster (total of 21 cores). For the comparative algorithm, more than 2 hours was required to achieve the same feat. In the proposed system, Hadoop’s distributed file system (HDFS) was used to achieve distributed storage while Apache Spark was used as the computing engine. The model development was based on a parallel model with full consideration of the high performance and throughput of distributed computing. Conclusively, the proposed HP-PL method can achieve good accuracy with less memory and time compared to the conventional methods of feature reduction. This tool can be publicly accessed at https://github.com/ahmed/Fast-HP-PL.
... The first pre-processing step of any face recognition system consists of relevant facial features extraction from original images. In order to extract these features, many techniques have been developed in recent years [1]. ...
Principal Component Analysis(PCA) is one of the key methods for solving data analysis problems with a large number of dimensions like face recognition. Nevertheless, classical PCA is based on Euclidean L2-norm which is very sensitive to noise and outliers. Recently, a new robust PCA approach has been proposed by replacing L2-norm with L1-norm (PCA-L1). However, PCA-L1 requires a lot of time to calculate the projection bases. To solve this problem, we propose to use a wavelets feature extraction method as pre-processing step to face recognition. Extensive experiments on two well-known face image datasets namely ORL and Georgia Tech Face Database(GTFD), show that the proposed approach minimizes execution time and has a recognition rate up to 96.7% for ORL and 85% for GTFD.
