Facial Recognition Using Square Diagonal Matrix Based on Two-Dimensional Linear Discriminant Analysis

Muntasa, Arif

doi:10.15866/irecos.v10i7.6623

Cited by 5 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The results of biometrics research have been implemented in many areas, including the corporate sector, defense and security, governance, and even in small gadgets. Biometrics have also been developed to obtain new breakthroughs, such as fingerprint recognition (Ito et al, 2006;Tan & Bhanu, 2006), facial detection (Yang et al, 2002;Rowley et al, 1998;Hjelmas & Low, 2001;Viola & Jones, 2004), automatic face-image data acquisition system (Wahab et al, 2015), and facial recognition (Muntasa, 2014;Muntasa, 2015;Shehzad et al, 2014). Facial detection plays an important role in facial recognition, as the results of facial detection will affect facial recognition such that an incorrect detection of a facial image will result in an incorrect facial recognition.…”

Section: Introductionmentioning

confidence: 99%

Windowing System Facial Detection based on Gabor Kernel Filter, Fast Fourier Transform, and Probabilistic Learning Vector Quantization

Muntasa¹

2017

IJTech

Self Cite

View full text Add to dashboard Cite

Facial detection is a crucial stage in the facial recognition process. Misclassification during the facial detection process will impact recognition results. In this research, windowing system facial detection using the Gabor kernel filter and the fast Fourier transform was proposed. The training set images, for both facial and non-facial images, were processed to obtain the local features by using the Gabor kernel filter and the fast Fourier transform. The local features were measured using probabilistic learning vector quantization. In this process, facial and non-facial features were classified using label 1 and -1. The proposed method was evaluated using facial and non-facial image testing sets, which were taken from the MIT+CMU image database. The testing images were enhanced first before the detection process using four different enhancement methods: histogram equalization, adaptive histogram equalization, contrast limited adaptive histogram equalization, and the single-scale retinex method. The detection results demonstrated that the highest average accuracy was 83.44%.

show abstract

Section: Introductionmentioning

confidence: 99%

Windowing System Facial Detection based on Gabor Kernel Filter, Fast Fourier Transform, and Probabilistic Learning Vector Quantization

Muntasa¹

2017

IJTech

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many approaches have been promoted to overcome the problems, i.e. Covariance-based subspace [1], Eigenface [2][3][4], Fisherface [5][6][7][8][9], Independent Component Analysis [10], Local manifold model [11][12], Laplacian Smoothing Transform (LST) [13][14], Kernel subspace [15], Tangent Space [16], Sparse Neighborhood [17], Wavelet [18], Graph embedding [19], Matrix-based Features [20], Vertical and Horizontal Information [21], Locally Linear Regression [22], and Homogeneous and Non-homogeneous Polynomial Model [23].…”

Section: Introductionmentioning

confidence: 99%

“…The LDA can preserve the sensitive information. Therefore, it can reduce the error rate when classification process is performed [5][6][7][8][9]. However, the LDA has also weaknesses, i.e., the correlation pattern was interpreted the similar from a group to each other, the function of discriminant built has a normal distribution for each group being compared.…”

Section: Introductionmentioning

confidence: 99%

Features Fusion based on the Fisherface and Simplification of the Laplacian Smoothing Transform

Muntasa¹

2017

ijeei

Self Cite

View full text Add to dashboard Cite

The newest model was proposed to extract the characteristic as distinctive attribute based on the Simplification of the Laplacian Smoothing Transform (S-LST) and the Fisherface. The proposed model is composed of two primary processes, i.e., training and testing processes. A training process involved features extraction based on the S-LST and the Fisherface, selection of the principal features, and features fusion of them. The proposed model has proved that the features can preserve the particular global information and the local feature areas of the rows and columns. The testing process is the stage to obtain the primary features through the new projection of the training sets. The last operation of the testing process is to measure the similarity of the features. The proposed model was tested using three different databases, i.e., the ORL, the YALE, and the UoB. The proposed model demonstrated the best performance, which is 100% recognition rate on the ORL and YALE when five images are implemented to the training sets. The maximum average of the recognition rate is 98.7% for the ORL database, 97.7% for the YALE database, and 94.53% for the UoB database. The comparison result showed that the proposed model performs better than the other approach, i.e., LST+LDA, S-MFA, S-NPE, S-LPP, S-LDA, LPP, LDA, and PCA.

show abstract

“…Principally, dimensionality reduction is used to obtain the main features of an object, so that computation time of the similarity measurement can be significantly reduced. Many methods have been developed to overcome high dimensionality, such as the Principal Component Analysis wellknown as PCA [7][8], Linear Discriminant Analysis [7], Kernel PCA [9][10], Linear Preserving Projection or Laplacianfaces [11][12], Gabor Wavelet [13], and Two-Dimensional Fisherface [14][15]. One of the most common and the oldest methods to extract the features is PCA.…”

Section: Introductionmentioning

confidence: 99%

Maximum Value of the Symmetrical Diagonal Matrix Based Two - Dimensional Eigenface s for Facial Recognition on the Variant Lighting and Expressions

Muntasa

Telang²

2016

ijeei

Self Cite

View full text Add to dashboard Cite

This research proposed a new approach to recognize face image under various lighting expression. The Proposed method is started by generating of the Left Diagonal Matrix (LDM) and the Right Diagonal Matrix (RDM). Subsequently, the dimensionality reduction is conducted by using Two-Dimensional Eigenface of the LDM and the RDM. The results of dimensionality reduction are selected the maximum value based on the corresponding features. Finally, the geometric similarity measurement model is carried out to obtain the recognition rate. The proposed method was evaluated using three different facial image databases, which are the YALE, the ORL and the UoB face image databases. Two until six column features were used to measure the similarity between the training and the testing sets. Experimental results revealed that the proposed method produced 92.2%, 94%, and 94.7% recognition rate for the YALE, the ORL, and the UoB face image databases respectively. The proposed method was also compared to the other methods. On the ORL face image database, the comparison results show that the proposed method outperformed to the Eigenface, the Fisherface, and the Laplacianfaces but not for O-Laplacianfaces and Two-Dimensional Principal Component Analysis method. On the YALE and the UoB face image databases, the proposed method outperformed to the other appearance methods which are the Eigenface, the Fisherface, the Laplacianfaces, the O-Laplacianfaces and Two-Dimensional Principal Component Analysis methods.In this research, Selection of the maximum values of the left and the right of the symmetrical diagonal matrix was proposed. To obtain the main features and to reduce the dimension and computation time, Two-Dimensional Principal Component Analysis was proposed to extract it. The proposed method ensured the features applied are the maximum value of the dominant features of the corresponding features.The rest of this paper was organized into four sections. the second section explains the weakness of one-dimensional of appearance methods. The third part presents the proposed method. The series of the proposed method are also presented in this part, which are the Left Diagonal Matrix or LDM, the Right Diagonal Matrix or RDM, Two-Dimensional Eigenface of the LDM, Two-Dimensional Eigenface of the RDM¸ selection of the maximum value of the LDM and the RDM features, and similarity measurement. Fourthly, experimental results are explained and followed by analysis. Finally, conclusions are written In the fifth section. The Weakness of One-Dimensional of Appearance Methods.In the biometrics field, there are three method models to extract the main features of an object, i.e. appearance-based, features-based, and hybrid-based. An appearance-based method extracted an object based on its vision. An object is viewed as space that consists of the features. Two crucial problems on an appearance-based method, which are high dimensional measured and computational time consumed. Principal Component Analysis (PCA or 1D-PCA) is the simplest method based o...

show abstract

Facial Recognition Using Square Diagonal Matrix Based on Two-Dimensional Linear Discriminant Analysis

Cited by 5 publications

References 0 publications

Windowing System Facial Detection based on Gabor Kernel Filter, Fast Fourier Transform, and Probabilistic Learning Vector Quantization

Windowing System Facial Detection based on Gabor Kernel Filter, Fast Fourier Transform, and Probabilistic Learning Vector Quantization

Features Fusion based on the Fisherface and Simplification of the Laplacian Smoothing Transform

Maximum Value of the Symmetrical Diagonal Matrix Based Two - Dimensional Eigenface s for Facial Recognition on the Variant Lighting and Expressions

Contact Info

Product

Resources

About