See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/361376462

Face Recognition Enhancement System by using Parallel Processing

Article in Guangdianzi Jiguang/Journal of Optoelectronics Laser · June 2022

CITATIONS READS

0 73

3 authors, including:

Ahmed Al-Araji
University of Technology, Iraq
105 PUBLICATIONS 508 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Modeling and Analysis of DC-DC Buck Converter for Mobile Applications View project

Deep Machine Learning for Arabic Sentiment Analysis View project

All content following this page was uploaded by Ahmed Al-Araji on 17 June 2022.

The user has requested enhancement of the downloaded file.

JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

Face Recognition Enhancement System by using Parallel Processing

*1 2 3
Rasha Khudiar Rija , Ghaida Muttasher & Ahmed Al-Araji
1,2,3
Computer Engineering Department University of Technology-Iraq, Baghdad-Iraq

ABSTRACT Nowadays, information security is a critical and complex issue. Security cameras are widely used at banks, ATMs,
airports, colleges, and workplaces, as well as everywhere there is a security system. One of the most difficult computer vision and
image processing tasks is face recognition. Faces are biometric systems that identify or authenticate a person by using a digital
image of that individual. In the security business, this system is employed. The facial recognition system must be capable of
recognizing functions by first extracting and then recognizing transitions (translation, rotation, and scaling), regardless of the
expression or lighting of the photograph), and aging (a challenging process). Several studies have been conducted to enhance face
recognition in terms of security, time, accuracy, etc. Time is a major factor that influences face recognition. Therefore, the
proposed system focuses on enhancing the results of execution time as well as the accuracy by using viola and Jones for detection,
DCT and PCA for feature extraction, and SVM for classification (recognition). The whole process of face recognition is done in
parallel processing to increase the speed of the system.

Keywords: Face Recognition, Face Detection, Feature Extraction, Parallel Processing.

I. INTRODUCTION
Face recognition has been chosen as an essential study issue in the realm of biometrics due to its non-contact and
non-stealing qualities. Face recognition technology has advanced significantly as machine learning, pattern
recognition, artificial intelligence, and computer vision technologies have advanced [1].Nowadays, recognition systems
play an essential role in future interactions between people and machines. Machines can complete tasks more
quickly, accurately, and securely. There are already reliable means of biometric personal identification, such as an iris
or fingerprint scanner. However, without the participant's involvement, facial recognition is frequently successful, the
aim today is to create a facial recognition system that is accurate, simple, fast and requires few processing resources.
The majority of face recognition algorithms use holistic aspects to represent the facial picture. The holistic traits are
extracted from the entire facial picture. [2].

Face recognition is a biometric identification system that uses facial characteristics to identify people. Face recognition
technology, also known as face recognition, comprises the purchase of a camera or camera with a face picture or
video stream, as well as the automatic detection and tracking of human faces in images. [3].Huge and high-
dimensional datasets are becoming an increasingly important component of modern life, from medical to economics,
and from social media to the Internet of Things (IoT). In a similar vein, due to their data-driven nature and capacity to
generalize to previously unknown data, machine learning algorithms are finding applications in every imaginable
domain. However, for these algorithms to be successful and economical, a significant amount of data preparation is
required. Dimension reduction and feature learning are important phases in this preprocessing because they
compress and extract relevant features from raw data that may be utilized in downstream machine learning algorithms
for classification, clustering, and so on [4].

Face detection and identification algorithms have been extensively explored and improved in terms of accuracy,
efficiency, and velocity throughout the preceding decade. High-performance computations are typically required by
face detection and identification methods. Microprocessor processing capability has increased significantly as
computing speed has improved. This increase, however, is insufficient to keep up with the fast improvements in image
and video technologies, particularly camera quality and real-time processing requirements. The system designers
used parallelism to implement concurrent processing concepts. Parallelism boosts processing performance by
increasing the number of hardware modules that may operate concurrently, resulting in the development of many
processes. To achieve high performance in future computer systems, parallel processing and multi-core architectures
will be important. The level of parallelization is specified in multi-core system standards, and performance is improved
by architectural details such as SIMD units or cache effect utilization [5].

II. FACE RECOGNITION

Face recognition has emerged as one of the most critical issues in computer vision and machine learning over the last
decade. Airport security, surveillance systems, automated education attainment, record management, laptop entertai
nment, virtual worlds, online banking, and video archiving all employ face recognition.. system of authentication, for
example, aim to identify a specific individual among enormous crowds with no physical interaction. Because the target
individual is compared against photographs from a database, the matching results do not require an expert to
understand. If known criminals' and terrorists' photos are maintained in a database, such an automated technique can
be utilized to discover them. Principle Component Analysis is one of the fundamental techniques for solving face
recognition difficulties (PCA) [6].

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

514
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

Face recognition is focused on identifying a person from a group of images. It is regarded as a superior biometric
technique to fingerprint, iris/retina, and speech recognition since it may be utilized in both active and inactive settings.
People are either standing in front of a camera or walking past one. However, employing face recognition is
challenging since faces are dynamic and can vary in emotion, lighting, position, and size.

Over the last two decades, computer vision researchers have paid increasing attention to face recognition. Many facial
recognition methods have been developed and may be classified. into three categories: holistic, feature-based, and
hybrid approaches. In fact, it is backed by its pervasiveness. Smart phones, laptops, and tablets are outfitted with
cameras. Furthermore, the picture quality from these cameras has improved. been improved in recent years. In
addition, Facebook and Google make use of powerful facial recognition techniques.

Holistic approaches employ full or partial picture pixels to represent global or local characteristics, whereas feature-
based methods use extracted facial features such as eyes, noses, and mouths to represent features. Then there are
hybrid techniques, which integrate both holistic and feature-based approaches. Some popular holistic-based
approaches are Eigenfaces, Fisherfaces, and local binary patterns (LBP) [7].

A few years ago, utilizing our fingerprints to unlock our smartphones was deemed the future. There is now a debate
on how facial recognition will be integrated into personal existence in the age of big data. Face detection may be
utilized to help with routine chores. A person may recognize faces by considering a variety of criteria, such as different
angles, hairstyles, markings on the face, the color of skin, and others, and attempting to match them with various
expressions.

The capacity to model faces in a multitude of styles, emotions, positions, lighting, and so on is required for the
digitalization task, and Fig.1 displays numerous stances for identifying a person's face [8].

Fig. 1 Face recognition in various positions

To create a reliable face recognition system, three main processes are followed: (1) detection of faces, (2) extraction
of features, and (3) face recognition [9] as shown in fig. 2. Because of the difficulty required in feature identification
and the dependency of recognition performance on feature detection accuracy, feature-based approaches are not
chosen for real-time FR. For real-time FR, holistic approaches are easier and utilize several feature extraction
techniques and classifiers. PCA has been widely utilized for the extraction of discriminating features, and
investigations have demonstrated that it outperforms Linear Discriminant Analysis (LDA) in realistic circumstances
[10].

Fig 2. Steps of Face Recognition System

Parallel computing has gained relevance in data science and related application fields because it is significantly more
suited than serial computing for studying, modeling, and evaluating complicated real-world scenarios. For example,
parallel computing has been successfully applied in a wide range of domains, including genomics, biotechnology,
geology, psychology, mechanical engineering, electrical circuits, microsystems, the environment, diagnostic imaging,
data mining, and big data analytics, to mention a few. Parallel processing (cores, computers, etc.) is the use of many
JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

515
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

computers' power at the same time to execute a computing job. In general, parallel computing is used when dealing
with circumstances that require the use of powerful computers. Due to the high resource requirements of face
recognition, the present suggested study will employ parallel processing with face recognition [12].

III. LITERATURE SURVEY

Many researchers are looking at biometric authentication solutions. Face recognition has become one of the most
extensively used issues in a variety of fields, including the use of password substitution in the launching of some apps
and other services [21].To satisfy high frame rate needs, Mahale et al. (2015) presented a modular hardware solution
for the algorithm. Because of the availability of vast external memory, the suggested architecture is scalable in terms
of the dimensions of the database and images. To reduce memory latencies, a new storing format on external
memory is used. The hardware solution's FPGA emulation can conduct face recognition on 128*128 images at (450)
recognitions per second using (450) face image classes. Because of its improved recognition algorithm and scalable
parallel architecture, the proposed FRS is well-suited for real-time surveillance applications as a result of its improved
recognition algorithm and widely spread design, but the system needs the extra hardware [10].

In Yao and Zhu's study (2016), they used SVM and Adaboost classifiers for face recognition, as well as the rapid PCA
approach to decrease the dimension of input face photos while retaining the majority of intrinsic facial characteristics.
The whole structure and each component are thoroughly covered. This study compared and tested the performance
of the two methods on facial photographs in diverse settings. In this research, PCA with SVM outperforms PCA with
Adaboost. All the results are positive, with the exception of the results on motion blur photos, which are negative. It is
possible that the characteristics are too complex for the classifiers to understand. It becomes difficult to discern
between facial representations [11].

Olmedo et al. (2017) developed a parallel approach for three tasks of the Viola-Jones face identification algorithm,
notably for the training step. The three tasks are as follows: integral image calculation, haar-like feature development,
and feature assessment. As a result, they suggested three parallel techniques for the face identification issue in
computer vision, which need a lot of computational power in real-time. The suggested solutions enhance OpenCV
implementation, according to their experimental findings. In reality, our technique resulted in shorter processing times
for the integral image calculation in virtually all circumstances and various picture sizes. Furthermore, the suggested
approach reduced timeframes in all experiments; all the results were achieved using CUDA only [12].

Yu Li and Xian Lin (2018) proposed a novel face recognition approach based on the Haar feature classifier, HOG
feature extraction, and fast-PCA dimension reduction. During the initial step of data preparation, the Haar feature
classifier is employed to extract backdrop interfering data. The HOG feature extraction method is then utilized to
recover the feature data from the face. The extracted data using the PCA technique is then compressed for usage in
data training and testing. Finally, the SVM approach is used to detect and identify the face. The system's recognition
rate must be increased [1].

Balachandran et al. (2019) used AI to create a functional facial recognition application. This study utilized the
VGGFace model in this neural net. Its application is divided into two parts: training and recognition. The training phase
involves adding new faces to the system, while the recognizing phase involves determining the identity of a face. The
program can run on several cores without issue. The results demonstrate that the program can scale successfully
over several cores when utilizing Tensorflow. When increasing the number of cores, the bottleneck may be caused
[13].

In Winarno et al.'s study (2019), they adopted an approach to the development of 2D to 3D image reconstruction
models utilizing convolutional neural networks (CNN) and the usage of PCA as the feature extraction technique. The
CNN technique is employed to generate a 3D face picture from a 2D face image. In the suggested facial recognition-
based attendance system, the PCA approach as a feature extraction method and the Mahalanobis method as a
classification method may both operate effectively. The proposed method can provide up to 98 percent accuracy in
facial recognition, but the system's execution time must be reduced [14].

Bhushan et al. (2022) suggested a new face identification technique that outperforms the DCT-SVM and PCA-SVM
models. It optimizes using a mix of DCT, SVM, and the Genetic Algorithm. Following the division of the dataset into
two pools for training and testing, as a consequence, the suggested model outperforms the DCT-SVM and PCA-SVM
models in terms of accuracy, training time, and classification time. The suggested approach was applied to databases
with size and type constraints. [15]

IV. THE PROPOSED ALGORITHM DESIGN AND IMPLEMENTATION

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

516
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

The proposed algorithm has two phases: training and testing. Images of (40) people in (10) different stances are
captured and preprocessed for coloured dye and picture scaling during the training phase. After preprocessing, the
Viola-Jones technique is utilize to determine the face region (the detection phase). Effective features are extracted
from accelerometer data (data compression) using DCT (discrete cosine transform). A PCA (Principal Component
Analysis) algorithm is used to minimize the dimensionality of the feature and perform feature extraction from
images in the DCT domain using a parallel computing approach (velocity). These characteristics are saved in the
database of information for the recognition of facial photos in various positions. In the testing phase, input pictures are
preprocessed, and recognized faces and multi-features are retrieved in the same way as they were in the training
phase. For face recognition using Support Vector Machine, these features are matched with the features contained in
the database of information (SVM),which is used to classify human activities. Next, the system determines whether or
not the customer is known (recognition phase) [8]. Python's programming language has useful functions for this
purpose. Computer vision and machine learning are applications that aid in the automatic processing of groups of
images [16]. The proposed methodology translates facial recognition techniques appropriate for large data
characteristics in the following ways and shown in fig 3.

Fig 3. The Proposed Algorithm For Face recognition system

Face Detection Technique by Haar Cascade

It is essentially the first step of a facial recognition system, and its objective is to locate and localize the face region
from the backdrop, which is usually congested and mixed with the face color, making it difficult to find the face. The
goal of face detection is to extract the face area from an image and precisely localize the face inside the picture. The
accuracy of face detection is dependent on the properties of the acquired image, such as noise and lighting
conditions. The following are some difficulties: Pose, facial expression, lighting, occlusion, and so forth. temporary
structural elements. In our proposed system, we use haar cascade for face detection [17].

The algorithm was devised by Paul Viola and Michael Jones. The system recognizes frontal faces better than sides,
above, or down-facing faces. Viola-Jones is competent in real-time. It locates the face in the colorful image after
recognizing it in the grayscale image. They had fundamentally rectangular characteristics, which are just one
component of a picture removed from another as shown in fig.4. Because their methodologies are machine learning-
based, they choose a collection of simple characteristics that they integrate into an effective, extensible classifier.

Their contribution was a rapid approach to computing these attributes and applying them to classifying faces. They
encountered an issue in that computing a huge group of pixels and summing them up is a time-consuming procedure
[18].

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

517
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

Fig 4. Haar Cascade Classifier

Feature Extraction Techniques

Facial recognition relies on feature extraction to capture the relevant components present in face space. The retrieved
characteristics must be useful for the subsequent processes, such as picture projection and subject categorization,
with an acceptable error rate. Furthermore, the feature extraction technique should be efficient in terms of computing
resources like memory and processing time. As a result, dimensional reduction techniques like DCT and PCA are
used in holistic-based methods [2].

DCT (Discrete Cosine Transform)

For both global and local face recognition systems, the discrete cosine transform (DCT) can be used. A discrete
cosine transform (DCT) is a transformation that describes a finite sequence of data as the sum of a series of cosine
functions with different frequencies. This method is extensively used in facial recognition systems, as well as in audio
and picture compression and spectrum techniques for numerically resolving differential equations [9].

PCA( Principal component analysis)

The PCA approach condenses a large number of possibly connected variables into a small number of incorrect
variables referred to as "principal components." PCA's goal is to decrease the enormous dimensionality of the data
space (observed variables) to the reduced intrinsic dimensionality of feature space (independent variables), which is
required to economically characterize the data. PCA computes the covariance matrix's Eigenvectors and projects the
original data onto a lower-dimensional feature space formed by Eigenvectors with high Eigenvalues. Eigenfaces are
Eigenvectors created by PCA that have been used to recognize and represent faces.

An image may additionally take into account the vector of dimension M N, resulting in a typical picture of 4 x 4
becoming a vector of dimension 16. Let the training set consist of X1, X2, X3, and XN photos. The set's average face
is defined as follows:
𝑁
1
𝑋 = ∑ 𝑋𝑖
𝑁
𝑖=1

To express the dispersion degree of all feature vectors associated with the average vector, compute the estimated
covariance matrix. The following is the definition of the covariance matrix Q:
𝑁
1
𝑄 = ∑(𝑋 − 𝑋𝑖)(𝑋 − 𝑋𝑖)𝑇
𝑁
𝑖=1

The Eigenvectors and Eigen-values are calculated using

𝐶𝑉 = 𝜆𝑉 (𝑉 ∈ 𝑅𝑛, 𝑉 ≠ 0)

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

518
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

Where V is the matrix Q's collection of eigenvectors associated with its eigenvalue λ. Project all of the person's
training pictures into the relevant Eigen-subspace:
𝑌𝐾 𝑖 = 𝑊 𝑇 (𝑋𝑖 ) (𝑖 = 1,2,3 … … 𝑁)

These are x projections and are referred to as principal components, commonly called eigenfaces. The facial pictures
are represented as a linear combination of the "principal components" of these vectors, as shown in fig. 5 [9].

Fig 5. An example of dimensional reduction using principal component analysis (PCA)

Face Recognition Technique by SVM ( Support Vector Machine)

Face recognition is at the heart of the recognition technology. A computer vision system that analyzes facial
characteristic data to establish identification is known as face recognition. There are two types of face recognition:
face detection and face recognition matching. Face recognition technology is based on a person's facial traits and a
facial photo or video stream that is delivered. Determine whether or not a human face exists, and if so, offer
information on the location, size, and position of each main facial function. The identifying attributes present in each
face are extracted and compared with existing faces to ascertain the identity of each face based on this data. The
technique of face recognition is done by an SVM classifier [11]. SVM is a machine-learning algorithm used to address
classification problems. By isolating data using an insulated plane known as a hyperplane, the technique tackles the
problem of data interference. The support vector (SV) is the collection of points nearest to the hyperplane. The
support vector distance from the hyperplane and classification accuracy are related in a favorable way.. For
attendance systems, SVM has been presented as the finest and most efficient algorithm [18].

V. RESULTS
The proposed system achieved good results when using DCT-PCA-SVM without parallel processing, but when using
parallel processing, the proposed system achieves better results in terms of execution time (Eigen face time, training
time, and recognition time) and recognition rate (accuracy), as shown in table1,fig 6 and fig 7.

Table 1. Results of applying the DCT-PCA-SVM in parallel processing and without parallel processing
Execution time(sec) without parallel processing With parallel processing
Eigen face time 0.615 0.414
Training time 0.032 0.027
Recognition time 0.032 0.027
Total time 0.648 0.438
Accuracy 91% 96%

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

519
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

Fig 6 The execution time of the proposed system Without parallel processing and with it.

Fig 7 The accuracy of the proposed system Without parallel processing and with it.

VI. CONCLUSION
Face recognition technology has long been a topic of study in the fields of pattern recognition and computer vision, as
a system that detects what sort of person a matched face is by automatically detecting and analyzing a face region in
an input still picture or video that includes a face. Face recognition is a type of identification recognition, comparable to
ID card recognition, fingerprint recognition, and iris recognition. The proposed system focuses on reducing the whole
system's execution time by using parallel processing with face recognition.. However, in real-time environments, time
is viewed as the most crucial problem. As a consequence, we concentrate on developing an enhanced face
recognition system by solving the time issue related to autonomous face recognition processing performance.
Accuracy is also an important factor that influences the performance of face recognition Therefore, the proposed
system achieved 91% accuracy without using parallel processing but achieved 96% when using parallel processing.
The proposed face recognition uses Haar cascade in the detection phase, DCT-PCA for data compression and
feature extraction, and SVM in the recognition phase.

REFERENCES
[1] X.Y. Li and Z.X. Lin, Face recognition based on HOG and fast PCA algorithm, In The Euro-China Conference on
Intelligent Data Analysis and Applications.(2018). Springer, Cham. 10-21.
[2] R.A.R. Ahmad, M.I. Ahmad, M.N.M. Isa and S.A. Anwar, Face recognition using assemble of low frequency of
DCT features, Bulletin of Electrical Engineering and Informatics, 8(2) 2019 541-550.
[3] Z. Yu, F. Liu, R. Liao, Y. Wang, H. Feng and X. Zhu, Improvement of face recognition algorithm based on neural
network. International Conference on Measuring Technology and Mechatronics Automation. (2018). IEEE. 229-
234.
[4] A. Gang ana W.U. Bajwa, FAST-PCA: A Fast and Exact Algorithm for Distributed Principal Component Analysis.
arXiv preprint arXiv:2108.12373. (2021). IEEE. 1-16.

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

520
JOURNAL OF OPTOELECTRONICS LASER ISSN:1005-0086

Volume 41 Issue 6, 2022

[5] Z. Mohmood, M.U.S. Khan, M. Jawad, S.U. Khan ana L.T, Yang, A parallel framework for object detection and
recognition for secure vehicle parking, In 2015 IEEE 17th International Conference on High Performance
Computing and Communications, 2015 IEEE 7th International Symposium on Cyberspace Safety and Security,
and 2015 IEEE 12th International Conference on Embedded Software and Systems. IEEE. 892-895.
[6] D.S. Ibrahim and S.Hamdy, Parallel architecture for face recognition using MPI. International Journal of Advanced
Computer Science and Applications (IJACSA).8(1) 2017 425-430.
[7] B.W. Yohanes, R.D. Airlangga, I. Setyawan, Real Time Face Recognition Comparison Using Fisherfaces and
Local Binary Pattern, International Conference on Science and Technology (ICST).(2018).IEEE. 1-5.
[8] S. Dalali and S.L, Face Recognition: Multi-features Extraction with Parallel Computation for Big Data, International
Conference on Computational Systems and Information Technology for Sustainable Solutions
(CSITSS).(2018).IEEE.133-144.
[9] Y. Kortli,M. Jrid, A. Al Falou and M. Atri, Face recognition systems: A survey. Sensors, 20(2) 2020. 1-36.
[10] G. Mahale,H. Mahale, A.Goel, S.K. Nandy, S. Bhattacharya and R. Narayan, Hardware solution for real-time face
recognition, International Conference on VLSI Design .(2015). IEEE, 81-86.
[11] M. Yao and C. Zhu, SVM and adaboost-based classifiers with fast PCA for face reocognition, International
Conference on Consumer Electronics-China (ICCE-China). (2016). IEEE,1-5.
[12] E. Olmedo, J. de la Calleja, A. Morales-Reyes, H.J. Escalante, A.B.U. Najera, M.A. Medina and A.B. Ruiz, A
parallel approach for the training stage of the Viola-Jones face detection algorithm, Intelligent Data Analysis.
21(5)2017. 1097-1115.
[13] B. Balachandran,K.F. Saad, K. Patel and N. Mekhiel, Parallel Computer For Face Recognition Using Artificial
Intelligence, International Conference on Computer Engineering and Systems (ICCES).(2019). IEEE. 158-162.
[14] E. Winarno, U.H. Al Amin, H. Februariyanti, P.W. Adi, w. Hadikurniawati and M.T. Anwar, Attendance system
based on face recognition system using cnn-pca method and real-time camera. International Seminar on
Research of Information Technology and Intelligent Systems (ISRITI).(2019). IEEE. 301-304.
[15] S. Bhushan, M. Alshehri, N. Agarwa, I. Keshta, J. Rajpurohit and A. Abugabah, A Novel Approach to Face
Pattern Analysis, Electronics.11(3) 2022.444.. 1-14.
[16] S. Laith. F.S. Taher, and A. A. Abdulrahman, Effectiveness of new algorithms for facial recognition based on deep
neural networks. International Journal of Nonlinear Analysis and Applications.13(1) 2022. 2171-2178.
[17] A.M. Al-bakry, S.O. Al-mamory, & H.H. Alfartosy, Real-Time Face Recognition System Using Software Agent. In
2017 Second Al-Sadiq International Conference on Multidisciplinary in IT and Communication Science and
Applications (AIC-MITCSA).(2017). IEEE. 236-241.
[18] J. Alghamdi, R. Alharthi, R. Alghamdi, W. Alsubaie, R. Alsubaie, D. Alqahtani,… and R. Alshammari, A survey on
face recognition algorithms. In 2020 3rd International Conference on Computer Applications & Information
Security (ICCAIS).(2020). IEEE.1-5.
[19] S.A. Alwahab, H.A. Nassr and B.J. Al-Khafaji, Designing a system for identifying persons using 3d images.
International Journal of Nonlinear Analysis and Applications.12(1) 2021. 2477-2481.
[20] R.K.Rija, G.Muttasher and A. Al-Araji, Payment Systems Based on Face Recognition: A Survey. Journal of
Optoelectronics Laser.41(5),2022. 563-571.
[21] A.S. Abdullah,M.A. Abed and I. Al Barazanchi, Improving face recognition by elman neural network using curvelet
transform and HSI color space. Periodicals of Engineering and Natural Sciences. 7(2) 2019.430-437

JOURNAL OF OPTOELECTRONICS LASER DOI: 10050086.2022.06.61

521

View publication stats