KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 11

Some of the metrics are blocked by your
consent settings
SELM: Siamese extreme learning machine with application to face biometrics
(2022-07-01)
Kudisthalert, Wasu
;
Pasupa, Kitsuchart
;
Morales, Aythami
;
Fierrez, Julian
Extreme learning machine (ELM) is a powerful classification method and is very competitive among existing classification methods. It is speedy at training. Nevertheless, it cannot perform face verification tasks properly because face verification tasks require the comparison of facial images of two individuals simultaneously and decide whether the two faces identify the same person. The ELM structure was not designed to feed two input data streams simultaneously. Thus, in 2-input scenarios, ELM methods are typically applied using concatenated inputs. However, this setup consumes two times more computational resources, and it is not optimized for recognition tasks where learning a separable distance metric is critical. For these reasons, we propose and develop a Siamese extreme learning machine (SELM). SELM was designed to be fed with two data streams in parallel simultaneously. It utilizes a dual-stream Siamese condition in the extra Siamese layer to transform the data before passing it to the hidden layer. Moreover, we propose a Gender-Ethnicity-dependent triplet feature exclusively trained on various specific demographic groups. This feature enables learning and extracting useful facial features of each group. Experiments were conducted to evaluate and compare the performances of SELM, ELM, and deep convolutional neural network (DCNN). The experimental results showed that the proposed feature could perform correct classification at 97.87 % accuracy and 99.45 % area under the curve (AUC). They also showed that using SELM in conjunction with the proposed feature provided 98.31 % accuracy and 99.72 % AUC. SELM outperformed the robust performances over the well-known DCNN and ELM methods.
Some of the metrics are blocked by your
consent settings
Weighted sparse representation using collaborative representation in kernel feature space based classification
(2021-08-01)
Matsushima, Kousuke
;
Matsusue, Mihoshi
;
Ruengprateepsang, Kavin
—In this paper, we proposed new method to deal non-linear information such as occlusion by using kernel feature space, which named Weighted Sparse Representation using collaborative representation in kernel feature space based classification (WSRCRKC). Kernel Sparse Representation-based Classification (KSRC) has shown good classification performance and robustness for the problem of nonlinear distribution of face images. To use locality information and eliminate the affection of luminance, Weighted Kernel Sparse Representation-based Classification (WKSRC) is proposed as an extension of KSRC by combining multiscale retinex algorithm. Furthermore, by making kernel Gram matrix sparse, we reduce the computation of face recognition. The experimental result shows that our proposal clearly improves the computational time while keeping accuracy high.
Some of the metrics are blocked by your
consent settings
Regional covariance matrix-based two-dimensional PCA for face recognition
(2020-01-01)
Titijaroonroj, Taravichet
;
Hancherngchai, Kangsadan
;
Rungrattanaubol, Jaratsri
Two-dimensional principal component analysis (2DPCA) is widely used in many applications, especially, face recognition. A key factor to improve the performance of the 2DPCA method comes from the efficiency of the covariance matrix. This paper believes that the effective eigenvector can be extracted when the effective covariance matrix is given. Therefore, the computing covariance matrix is a focus point in this paper. The set of the covariance matrix in the 2DPCA and its extensions is usually represented with a single directional correlation, which is then used to obtain a mean covariance matrix by using the average technique. This causes in obtaining the ineffective eigenvector since the covariance matrix is ineffective. In order to obtain the effective eigenvector, a regional covariance matrix-based on 2DPCA method (RCM-2DPCA) is proposed here. The contribution of this paper consists of two main parts including (i) regional matrix calculation for computing the two directional correlations and (ii) ELSSP conversion for extracting the effective representation of the covariance matrix. The experimental results show that the performance of the proposed method is higher than the baseline methods including 2DPCA, I-2DPCA, Bi2DPCA, 2D2PCA and ILM-2DPCA methods on a basis of three well-known datasets-ORL Face, Yale Face, and Yale Face extended B+ datasets.
Some of the metrics are blocked by your
consent settings
Biometrics based on facial landmark with application in person identification
(2019-01-01)
Juhong, Aniwat
;
Purahong, Boonchana
;
Suwan, Supakorn
;
Pitavirooj, Chuchart
This paper presents a novel technique for face recognition based on facial landmarks extracted automatically. Our landmarks are those associated with eyes mouth and nose. To extract facial landmarks, we first use Haar cascade algorithm to detect the face ROI following by Haar cascade algorithm for the eye, mouth and nose ROI determination. To find landmark associated with the eye, we convert eye ROI image to binary image using thresholding algorithm. To exclude the eyebrow region, we apply horizontal radon transform. The project data will then be used to separate the eyebrow region from the eye region. To detect eye-related landmark, vertical radon transform is applied. With the vertical projection data, the outermost pixel can be identified and the associated eye landmark can be determined. The similar technique can then be used to identify landmarks associated with the nose and mouth area. Given the correspond landmarks on the reference face and the query face, geometric transformation can be determined using normal equation bases on minimized mean squared error. The two faces are then aligned. To provide the quantitative measurement, the two aligned face are converted to edge image using canny edge algorithm. The distance map error between the two aligned edge facial images is then used to identify the query face. The purposed algorithm for person identification was tested on the face database resulting in a very high accuracy.
Some of the metrics are blocked by your
consent settings
Face recognition based on facial landmark detection
(2017-12-19)
Juhong, Aniwat
;
Pintavirooj, C.
This paper presents a novel technique for face recognition based on facial landmarks extracted automatically. Our landmarks are those associated with eyes mouth and nose. With the extracted landmarks, the area triplets and the associated geometric invariance are formed. We opt to use area and triangle confined within the triangle as the invariance. To bypass the perspective constraints, we take the face image with high focal length and at the farther distance. Orthogonal projection and Euclidean transformation are then assumed. As area is relative invariance under Euclidean transformation, the absolute area ratios between consecutive area triples are applied. Our purposed algorithm is tested successfully to identify person and could be a promising technique for facial recognition.
Some of the metrics are blocked by your
consent settings
An evaluation of face recognition algorithms and accuracy based on video in unconstrained factors
(2017-04-05)
Jaturawat, Phichaya
;
Phankokkruad, Manop
Face recognition is the biometric personal identification that gaining a lot of attention recently. This method has the ability to identify a person from still image and video by using human face. For the accurate recognition, algorithm and reference database needs to be concerned. However, in the practical system have many external factors that affect to the recognition accuracy differently for each algorithm. This is a challenge problem of class attendance recording system deployment, which has uncontrolled environments. This paper comparing three well known algorithm that are Eigenfaces, Fisherfaces, and LBPH by adopts our new database that contains a face of individuals with variety of pose and expression. The experiment of face recognition in video conducted by varied the external factors that are light exposure, noise, and the video resolution, in the possible range. The results showed LBPH got the highest accuracy in all experiments, but this algorithm has the higher impact of the negative light exposure and high noise level more than the others that are statistical approach.
Some of the metrics are blocked by your
consent settings
Impact of facial expressions and posture variations in face recognition rate on different image databases
(2017-01-01)
Jaturawat, Phichaya
;
Phankokkruad, Manop
In this paper, the impact of facial expression and posture variations in face recognition were studied by using three face recognition algorithms that are Eigenfaces, Fisherfaces, and LBPH in terms of recognition accuracy. In order to find the type of algorithms that works efficiently for face recognition in video. The experiment was conducted by using two different databases with three amounts of image in training set. DB-one is uncontrolled people in the images, and DB-two is controlled facial expressions and posture. The results show the facial expression and posture variations have a lot of impact to Eigenfaces and Fisherfaces and the LBPH got the impact less than the others. It concluded that the impact of facial expression and postures are different on each algorithm, and impacted to the recognition accuracy.
Some of the metrics are blocked by your
consent settings
Multi-pipeline architecture for face recognition on FPGA
(2009-11-18)
Visakhasart, Sathaporn
;
Chitsobhuk, Orachat
In this paper, a new multi-pipeline architecture is proposed for face recognition system on FPGA. The proposed structure consists of four main units: Multi-Pipeline Control Unit (MPCU), Process Element Unit (PEU), Region Summing Unit (RSU), and Recognition Indexing Unit (RIU). Four recognition techniques: Principal Component Analysis (PCA), Modular PCA (MPCA), Weight MPCA (WMPCA), and Wavelet based techniques are adopted to evaluate the efficiency of the proposed architecture using several standard face databases. The experimental results show that the proposed architecture helps minimizing processing time through its multi-pipeline processes while still maintains high recognition rate. Moreover, the design has encouraged the reduction in hardware resources by utilizing the proposed reusable modules. © 2009 IEEE.
Some of the metrics are blocked by your
consent settings
Relevance-weighted (2D)2LDA image projection technique for face recognition
(2009-08-01)
Sanayha, Waiyawut
;
Rangsanseri, Yuttapong
In this paper, a novel image projection technique for face recognition application is proposed which is based on linear discriminant analysis (LDA) combined with the relevance-weighted (RW) method. The projection is performed through 2-directional and 2-dimensional LDA, or (2D)2LDA, which simultaneously works in row and column directions to solve the small sample size problem. Moreover, a weighted discriminant hyperplane is used in the between-class scatter matrix, and an RW method is used in the within-class scatter matrix to weigh the information to resolve confusable data in these classes. This technique is called the relevance-weighted (2D)2LDA, or RW(2D) 2LDA, which is used for a more accurate discriminant decision than that produced by the conventional LDA or 2DLDA. The proposed technique has been successfully tested on four face databases. Experimental results indicate that the proposed RW(2D)2LDA algorithm is more computationally efficient than the conventional algorithms because it has fewer features and faster times. It can also improve performance and has a maximum recognition rate of over 97%. © 2009.
Some of the metrics are blocked by your
consent settings
Weighted LDA image projection technique for face recognition
(2009-01-01)
Sanayha, Waiyawut
;
Rangsanseri, Yuttapong
In this paper, we propose a novel image projection technique for face recognition applications based on Fisher Linear Discriminant Analysis (LDA). The projection is performed through a couple subspace analysis for overcoming the "small sample size" problem. Also, weighted pairwise discriminant hyperplanes are used in order to provide a more accurate discriminant decision than that produced by the conventional LDA. The proposed technique has been successfully tested on three face databases. Experimental results indicate that the proposed algorithm outperforms the conventional algorithms. Copyright © 2009 The Institute of Electronics, Information and Communication Engineers.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results