Linear Discriminant Analysis Feature Research Articles

Two Proposed Models for Face Recognition: Achieving High Accuracy and Speed with Artificial Intelligence

In light of the development in computer science and modern technologies, the impersonation crime rate has increased. Consequently, face recognition technology and biometric systems have been employed for security purposes in a variety of applications including human-computer interaction, surveillance systems, etc. Building an advanced sophisticated model to tackle impersonation-related crimes is essential. This study proposes classification Machine Learning (ML) and Deep Learning (DL) models, utilizing Viola-Jones, Linear Discriminant Analysis (LDA), Mutual Information (MI), and Analysis of Variance (ANOVA) techniques. The two proposed facial classification systems are J48 with LDA feature extraction method as input, and a one-dimensional Convolutional Neural Network Hybrid Model (1D-CNNHM). The MUCT database was considered for training and evaluation. The performance, in terms of classification, of the J48 model reached 96.01% accuracy whereas the DL model that merged LDA with MI and ANOVA reached 100% accuracy. Comparing the proposed models with other works reflects that they are performing very well, with high accuracy and low processing time.

Deep and Shallow Feature Fusion in Feature Score Level for Palmprint Recognition

Contactless palmprint recognition offers friendly customer experience due to its ability to operate without touching the recognition device under rigid constrained conditions. Recent palmprint recognition methods have shown promising accuracy; however, there still exist some issues that need to be further studied such as the limited discrimination of the single feature and how to effectively fuse deep features and shallow features. In this paper, deep features and shallow features are integrated into a unified framework using feature‐level and score‐level fusion methods. Specifically, deep feature is extracted by residual neural network (ResNet), and shallow features are extracted by principal component analysis (PCA), linear discriminant analysis (LDA), and competitive coding (CompCode). In feature‐level fusion stage, ResNet feature and PCA feature are dimensionally reduced and fused by canonical correlation analysis technique to achieve the fused feature for the next stage. In score‐level fusion stage, score information is embedded in the fused feature, LDA feature, and CompCode feature to obtain a more reliable and robust recognition performance. The proposed method achieves competitive performance on Tongji dataset and demonstrates more satisfying generalization capabilities on IITD and CASIA datasets. Comprehensive validation across three palmprint datasets confirms the effectiveness of our proposed deep and shallow feature fusion approach.

Performance Evaluation of Block-Sized Algorithms for Majority Vote in Facial Recognition

Facial recognition (FR) is a pattern recognition problem, in which images can be considered as a matrix of pixels.There are manychallenges that affect the performance of face recognitionincluding illumination variation, occlusion, and blurring. In this paper,a few preprocessing techniques are suggested to handle the illumination variationsproblem. Also, other phases of face recognition problems like feature extraction and classification are discussed. Preprocessing techniques like Histogram Equalization (HE), Gamma Intensity Correction (GIC), and Regional Histogram Equalization (RHE) are tested inthe AT&T database. For feature extraction, methods such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Independent Component Analysis (ICA), and Local Binary Pattern (LBP) are applied. Support Vector Machine (SVM) is used as the classifier. Both holistic and block-based methods are tested using the AT&T database. For twelve different combinations of preprocessing, feature extraction, and classification methods, experiments involving various block sizes are conducted to assess the computation performance and recognition accuracy for the AT&T dataset.Using the block-based method, 100% accuracy is achieved with the combination of GIC preprocessing, LDA feature extraction,and SVM classification using 2x2 block-sizingwhile the holistic method yields the maximum accuracy of 93.5%. The block-sized algorithm performs better than the holistic approach under poor lighting conditions.SVM Radial Basis Function performs extremely well on theAT&Tdataset for both holistic and block-based approaches.

Instant learning based on deep neural network with linear discriminant analysis features extraction for accurate iris recognition system

Instant learning based on deep neural network with linear discriminant analysis features extraction for accurate iris recognition system

An improved intelligent water drops feature selection for finger vein recognition

In recent years, finger vein recognition has gained a lot of attention and been considered as a possible biometric feature. Various feature selection techniques were investigated for intrinsic finger vein recognition on single feature extraction, but their computational cost remains undesirable. However, the retrieved features from the finger vein pattern are massive and include a lot of redundancy. By using fusion methods on feature extraction approaches involving weighted averages, the error rate is minimized to produce an ideal weight. In this research, a novel combinational model of intelligent water droplets is proposed along with hybrid PCA LDA feature extraction for improved finger vein pattern recognition. Initially, finger vein images are pre-processed to remove noise and improve image quality. For feature extraction, Linear Discriminant Analysis (LDA) and Principle Component Analysis (PCA) are employed to identify the most relevant characteristics. The PCA and LDA algorithms combine features to accomplish feature fusion. A global best selection method using intelligent water drops (GBS-IWD) is employed to find the ideal characteristics for vein recognition. The K Nearest Neighbour Classifier was used to recognize finger veins based on the selected optimum features. Based on empirical data, the proposed method decreases the equal error rate by 0.13% in comparison to existing CNN, 3DFM, and JAFVNet techniques. The overall accuracy of the proposed GBSPSO-KNN is 3.89% and 0.85% better than FFF and GWO, whereas, the proposed GBSIWD-KNN is 4.37% and 1.35% better than FFF and GWO respectively.

Image Classification of Tempe Fermentation Maturity Using Naïve Bayes Based on Linear Discriminant Analysis

One of the foods in Indonesia that has a lot of nutritional content and benefits, one of which is tempeh. Tempe is usually made by fermenting soybeans with mold under special conditions to become tempeh. In the fermentation process, tempeh producers need to monitor the maturity of the tempeh until it is suitable for consumption. To detect this maturity requires a separate effort, so that an image processing approach is proposed in this study with the support of feature selection. An image allows for various features to be taken, such as texture features using GLCM and various color features including RGB, HSV, LAB, CMYK, YUV, HCL, HIS, LCH. With so many features, it is necessary to do a selection so that computation in its classification becomes efficient. This study aims to classify tempeh fermented images using the Naive Bayes method with Linear Discriminant Analysis (LDA)feature selection for GLCM features and eight color features. Tempe fermentation image is divided into three classes, namely raw, ripe and rotten. Based on the experimental results, the average accuracy in the test is 84.06%. In testing the fastest time is 1.87 seconds and the longest is 2.20 seconds. This shows that the classification of fermented tempeh maturity with Naive Bayes with LDA feature selection can work well.

Adhesive bond quality classification using machine learning algorithms based on ultrasonic pulse-echo immersion data

In this study, we explored the detection of weak bonds (WBs) due to contamination and faulty curing (FC) using linear ultrasound and machine learning. For this purpose, aluminium single-lap adhesive joints containing three variants of bonding quality were investigated: perfect bond, WB due to release agent (RA) contamination, and WB due to FC. The data, according to the deviation of the bonding protocol, were arranged in two groups, creating two datasets: distinct and complete. Each dataset included all bonding conditions (perfect, RA, and FC), although the distinct dataset contained only marginal cases, which were expected to be well separable, whereas the complete dataset included data with minor deviations from the bonding protocol. Pulse-echo C-scan images were acquired for all prepared samples in the immersion tank, and 45 features were initially extracted from the time traces representing each bonding group. The initial data were analysed via a t-test and pairwise correlation analysis to reveal statistically significant features. Then, we performed dimensionality reduction using tree-based, recursive, sequential, and linear discriminant analysis (LDA) feature selectors to explore feature importance and classification accuracy with different feature subsets. Finally, the important features identified with the different feature selectors were fed to support vector machine (SVM) classifiers, and the classification accuracies were compared amongst the different feature subsets. The classification accuracy using a distinct dataset in some cases demonstrated nearly 99% accuracy, indicating that significant bonding protocol deviations could be easily detected. It was demonstrated that classification accuracy increased with the number of features. However, even in the case of the 2D feature space obtained using linear discriminant analysis, the bonding quality classification accuracy remained higher than 84%. The feature subspace reduction with LDA demonstrated sufficient classification accuracy and an improvement of nearly 40% in training time compared with that for the initial feature set. Thus, the classical ultrasonic pulse-echo C-scan with an LDA feature transformation and SVM classifier could be used to identify the deviations in the bonding protocol in aluminium single-lap adhesive joints.

Facial Expression Recognition Based on LDA Feature Space Optimization.

With the development of artificial intelligence, facial expression recognition has become an important part of the current research due to its wide application potential. However, the qualities of the face features will directly affect the accuracy of the model. Based on the KDEF face public dataset, the author conducts a comprehensive analysis of the effect of linear discriminant analysis (LDA) dimensionality reduction on facial expression recognition. First, the features of face images are extracted respectively by manual method and deep learning method, which constitute 35-dimensional artificial features, 128-dimensional deep features, and the hybrid features. Second, LDA is used to reduce the dimensionality of the three feature sets. Then, machine learning models, such as Naive Bayes and decision tree, are used to analyze the results of facial expression recognition before and after LDA feature dimensionality reduction. Finally, the effects of several classical feature reduction methods on the effectiveness of facial expression recognition are evaluated. The results show that after the LDA feature dimensionality reduction being used, the facial expression recognition based on these three feature sets is improved to a certain extent, which indicates the good effect of LDA in reducing feature redundancy.

Early Detection of Breast Cancer Tumors using Linear Discriminant Analysis Feature Selection with Different Machine Learning Classification Methods

Globally, the frequency of breast cancer and its morality speak to a critical and developing risk for the developing countries. In Asia, Pakistan has the biggest rate of breast cancer. It is evaluated that every year 83,000 cases were reported in Pakistan and over 40,000 deaths are caused by breast cancer. Patients suffering from this malignancy have a better chance of surviving if they are diagnosed early. Many Early identification of breast cancer can be achieved using data mining techniques, allowing preventative treatments to be done. In this research Wisconsin Breast Cancer Dataset (WBCD) and Duke Breast cancer dataset (DBDS) are used with Linear Discriminant Analysis (LDA) feature selection with Support Vector Machine (SVM), Decision Tree (DT), Neural Network and Random Forest (RF) machine learning classifiers to predict breast cancer tumors. The finding of the proposed model is that feature selections through LDA improve the accuracy of detecting tumors and also reduce time duration of executing model. The best machine learning model with LDA feature selection is Neural Network Model with highest accuracy 1.00 among all classification models and also consume less time.

Early Detection of Breast Cancer Tumors Using Linear Discriminant Analysis Feature Selection with Different Classification Methods

Globally, the frequency of breast cancer and its morality speaks to a critical and developing risk for developing countries. In Asia, Pakistan has the biggest rate of breast cancer. It is evaluated that every year 83,000 cases were reported in Pakistan and over 40,000 deaths are caused by breast cancer. Many Early identifications of breast cancer can be achieved using data mining techniques, allowing preventative treatments to be done. Patients suffering from this malignancy have a better chance of surviving if they are diagnosed early. So, there is a need to develop a system to diagnose breast cancer tumors at an early stage. In this research WBCD and Duke Breast cancer datasets (DBDS) are used with Linear Discriminant Analysis (LDA) feature selection. Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF) are classifiers to predict breast cancer tumors.

Integrating Quantitative Radiomics in De-intensification Treatment for Oropharyngeal Carcinoma

Deintensification for early-stage oropharyngeal carcinoma includes surgery +/- adjuvant radiation therapy without the use of concurrent chemotherapy. Pathologic findings of nodal extracapsular extension (ECE), however, obligate the use of postoperative concurrent chemotherapy in these patients. The purpose of this study was to develop a robust predictive model for ECE using quantitative radiomics.Between 2016 and 2020, 81 patients with newly diagnosed resectable oropharyngeal carcinoma underwent upfront transoral robotic surgery and neck dissection at our institutions. In all these patients, high-resolution CT scans were done preoperatively and were concluded to have no evidence of ECE by experienced head and neck radiologists. Eight five percent of tumors were HPV-positive. Median age was 63 years. Radiomics features (107 features) from the primary and extended nodal regions (1-3 mm extension) were extracted to generate predictive models. Two feature selection approaches were evaluated: Minimum Redundancy Maximum Relevance (MRMR) and Linear Discriminant Analysis (LDA). Logistic regression, decision tree, and random forests classifier were chosen for our analysis.Thirty eight percent of patients were found to have ECE in at least one of the nodes, and therefore received concurrent chemoradiation postoperatively. There was a total of 119 pathological positive nodes. Of these positive nodes, 31 nodes were found to have ECE with a ECE of less than 1 to 10mm. Five-fold stratified cross-validation was used for evaluation. A 96% accuracy (0.95 AUC, 0.96 recall) was achieved from the primary node without extension, using LDA feature selection. An improved result of 97% accuracy (0.96 AUC, 0.97 recall) was obtained with the 3mm extension. MRMR attained the maximum accuracy of 77% (0.64 AUC, 0.36 recall) with the 3mm extension.Quantitative radiomics allows a robust prediction of ECE, thereby optimizing de-intensification therapy for oropharyngeal carcinoma.

Impedimetric Analysis of Trabecular Bone Based on Cole and Linear Discriminant Analysis

A spatially unambiguous characterization of electrical properties of osseous tissues is important for the therapy of osteopathy via electrical stimulation. Accordingly, the study aimed to characterize the highly inhomogeneous composition and structures of different anatomical regions of trabecular bone based on their electrical properties. The electrical properties of 64 porcine trabecular bone samples were analyzed in a parallel plate electrode configuration and compared with published results. Therefore, a novel method, combining traditional Cole model with a linear discriminant analysis (LDA), was developed to discriminate the different regions, i.e., femur head, greater trochanter, and femur neck. Possible mechanisms behind the distinction for different regions could be interpreted from both methods. Respective adjacent regions with similar structure and composition could be distinguished from statistically significant differences of Cole parameters, i.e., α (p < 0.01) and R∞ (p < 0.05). The latter was correlated especially with water content, indicating an association of individual differences in microstructures in particular with conductivity. Conversely, different regions were unambiguously discriminated with LDA based on permittivity or conductivity. Contributions to the discrimination were explicitly reflected by the coefficients of the derived LDA features. A clear distinction was obtained especially for a frequency response at 950 kHz. Moreover, predictions for the classification of unspecified samples assigned them correctly to their origin with a success of 92.9%. The combination of both methods offers the possibility for a spatially resolved and eventually patient specific discrimination and evaluation of bone tissues and their response to therapies, notably electrical stimulation.

Emotion Recognition of EEG Signals Based on the Ensemble Learning Method: AdaBoost

In recent years, with the continuous development of artificial intelligence and brain-computer interface technology, emotion recognition based on physiological signals, especially, electroencephalogram (EEG) signals, has become a popular research topic and attracted wide attention. However, how to extract effective features from EEG signals and accurately recognize them by classifiers have also become an increasingly important task. Therefore, in this paper, we propose an emotion recognition method of EEG signals based on the ensemble learning method, AdaBoost. First, we consider the time domain, time-frequency domain, and nonlinear features related to emotion, extract them from the preprocessed EEG signals, and fuse the features into an eigenvector matrix. Then, the linear discriminant analysis feature selection method is used to reduce the dimensionality of the features. Next, we use the optimized feature sets and train a classifier based on the ensemble learning method, AdaBoost, for binary classification. Finally, the proposed method has been tested in the DEAP data set on four emotional dimensions: valence, arousal, dominance, and liking. The proposed method is proved to be effective in emotion recognition, and the best average accuracy rate can reach up to 88.70% on the dominance dimension. Compared with other existing methods, the performance of the proposed method is significantly improved.

Deteksi kavitasi menggunakan linear discriminant analysis pada pompa sentrifugal

Cavitation is a phenomenon that often occurs in the centrifugal pumps. The impact of cavitation is a decrease in pump performance which will affect the ongoing production process in the industries. It is important to have a method to detect the phenomenon of cavitation early. The vibration signal is a parameter that is often used in detecting cavitation or other faulty components. One of the methods is based on the pattern recognition i.e. machine learning. Linear Discriminant Analysis (LDA) is a machine learning algorithm that has the advantage of reducing the parameters used into low dimensions without reducing the accuracy of their classification. The study proposes LDA to classify normal conditions, initial cavitation, intermediate cavitation and severe cavitation. The recording of the vibration signal is taken using the an accelerometer mounted on the inlet of the centrifugal pump. The vibration signal is then extracted using 10 statistic parameters of time domain as the LDA feature selection, namely mean, RMS, standard deviation, kurtosis, skewness, crest factor, clearance factor, shape factor, variance and peak value. The results shows that the LDA classifier can detect and classify cavitation conditions with an accuracy rate of 98.8% on training and 99.6% on testing. The shape factor, kurtosis, skewness and RMS parameters are a combination of parameters that have a large contribution to the classifier to detect and classify cavitation conditions.Keywords: Linear Discriminant Analysis (LDA), cavitation, centrifugal pump, statistical parameter

Hybrid CNN assisted Computer Aided Diagnosis System for Glaucoma Detection and Classification: GlaucoNet+

The exponential rise in technologies has revitalized academia-industries to achieve more efficient computer aided diagnosis systems. It becomes inevitable especially for Glaucoma detection which has been increasing with vast pace globally. Most of the existing approaches employs morphological features like optical disk and optical cup information, optical cup to disk ratio etc; however enabling optimal detection of such traits has always been challenge for researchers. On the other hand, in the last few years deep learning methods have gained widespread attention due to its ability to exploit fine grained features of images to make optimal classification decision. However, reliance of such methods predominantly depends on the presence of deep features demanding suitable feature extraction method. To achieve it major existing approaches extracts full-image features that with high dimensional kernel generates gigantically huge features, making classification computationally overburdened. Therefore, retaining optimal balance between deep features and computational overhead is of utmost significance for glaucoma detection and classification. With this motive, in this paper a novel hybrid deep learning model has been developed for Glaucoma detection and classification. The proposed Hybrid CNN model embodies Stacked Auto-Encoder (SAE) with transferable learning model AlexNet that extracts high dimensional features to make further two-class classification. To achieve computational efficiency, In addition to the classical ReLu and dropout (50%), we used Principle Component Analysis (PCA) and Linear Discriminant Analysis (LDA) algorithms. We applied 10-fold cross validation assisted Support Vector Machine classifier to perform two-class classification; Glaucomatous and Normal fundus images. Simulation results affirmed that the proposed Hybrid deep learning model with LDA feature selection and SVM-Poly classification achieves the maximum accuracy of 98.8%, precision 97.5%, recall 97.5% and F-Measure of 97.8%.

Analysis of Facial Image Extraction on Facial Recognition using Kohonen SOM for UNPRI SIAKAD Online User Authentication

Academic Information System (Sistem Informasi Akademik aka SIAKAD) Online of Universitas Prima Indonesia (UNPRI) is one of the applications used to facilitate the administration process of lectures which includes the filling process of study plan cards (Kartu Rencana Studi aka KRS), study result cards (Kartu Hasil Studi aka KHS), class schedules, submission of research titles, seminars, and other processes. SIAKAD UNPRI can be accessed by students, lecturers, and academics where every user has a password that has been encrypted to maintain the security of information from people who are not responsible, password security using the encryption method needs to be changed regularly, but there are still many students, lecturers and academic community who are reluctant to change passwords. To improve the security verification stage for SIAKAD users, we propose a face recognition feature approach. Face recognition is a feature that allows the identification of someone from a digital image or video. The way the facial recognition method works is by comparing face data from the camera or images with images that were previously stored in a database. In this study, the Kohonen SOM method is proposed for face identification based on the feature extraction approach of discrete cosine transform (DCT), linear discriminant analysis (LDA) and principal component analysis (PCA) to improve the security of UNPRI SIAKAD users. The analytical framework is done by requiring students to do face taking, where each student will save 5 (five) faces extracted with facial features using the DCT, LDA and PCA model approach, feature extraction results are used as input to the Kohonen SOM network for training and testing facial recognition, then analysis of the effect of DCT, LDA and PCA feature extraction on the Kohonen network on facial recognition accuracy.

Local Wavelet Acoustic Pattern: A Novel Time–Frequency Descriptor for Birdsong Recognition

Investigating the identity, distribution, and evolution of bird species is important for both biodiversity assessment and environmental conservation. The discrete wavelet transform (DWT) has been widely exploited to extract time–frequency features for acoustic signal analysis. Traditional approaches usually compute statistical measures (e.g., maximum, mean, standard deviation) of the DWT coefficients in each subband independently to yield the feature descriptor, without considering the intersubband correlation. A new acoustic descriptor, called the local wavelet acoustic pattern (LWAP), is proposed to characterize the correlation of the DWT coefficients in different subbands for birdsong recognition. First, we divide a variable-length birdsong segment into a number of fixed-duration texture windows. For each texture window, several LWAP descriptors are extracted. The vector of locally aggregated descriptors (VLAD) is then used to aggregate the set of LWAP descriptors into a single VLAD vector. Finally, principal component analysis (PCA) plus linear discriminant analysis (LDA) are employed to reduce the feature dimensionality for classification purposes. Experiments on two birdsong datasets show that the proposed LWAP descriptor outperforms other local descriptors, including linear predictive coding cepstral coefficients, Mel-frequency cepstral coefficients, perceptual linear prediction cepstral coefficients, chroma features, and prosody features. Furthermore, the proposed LWAP descriptor, followed by VLAD encoding, PCA plus LDA feature extraction, and a simple distance-based classifier, yields promising results that are competitive with those obtained by the state-of-the-art convolutional neural networks.

Deep-learning-based automatic computer-aided diagnosis system for diabetic retinopathy.

The high-pace rise in advanced computing and imaging systems has given rise to a new research dimension called computer-aided diagnosis (CAD) system for various biomedical purposes. CAD-based diabetic retinopathy (DR) can be of paramount significance to enable early disease detection and diagnosis decision. Considering the robustness of deep neural networks (DNNs) to solve highly intricate classification problems, in this paper, AlexNet DNN, which functions on the basis of convolutional neural network (CNN), has been applied to enable an optimal DR CAD solution. The DR model applies a multilevel optimization measure that incorporates pre-processing, adaptive-learning-based Gaussian mixture model (GMM)-based concept region segmentation, connected component-analysis-based region of interest (ROI) localization, AlexNet DNN-based highly dimensional feature extraction, principle component analysis (PCA)- and linear discriminant analysis (LDA)-based feature selection, and support-vector-machine-based classification to ensure optimal five-class DR classification. The simulation results with standard KAGGLE fundus datasets reveal that the proposed AlexNet DNN-based DR exhibits a better performance with LDA feature selection, where it exhibits a DR classification accuracy of 97.93% with FC7 features, whereas with PCA, it shows 95.26% accuracy. Comparative analysis with spatial invariant feature transform (SIFT) technique (accuracy-94.40%) based DR feature extraction also confirms that AlexNet DNN-based DR outperforms SIFT-based DR.

Accurate Human Recognition by Score-Level and Feature-Level Fusion Using Palm–Phalanges Print

Here, a highly precise human identification system is developed using a newly proposed biometric trait—palm–phalanges print (PPP). For PPP, NSIT database has been used which includes palm–phalanges. This database consists of anterior hand images of fifty individuals with ten samples each. To crop the region of interest from hand samples to get palm–phalanges, database is preprocessed. First, it has been shown that each finger phalange can be used as a biometric modality and give moderate/sufficient performance for low-accuracy system. For feature extraction, histograms of oriented gradients, GMF feature, mean and AAD methods have been used. To further enhance the performance, score-level and feature-level fusion strategies are applied and compared. Score-level fusion is performed using different fusion rules. Next for feature-level fusion, five methods are used: (1) simple concatenate, (2) PCA feature fusion, (3) linear discriminant analysis feature fusion, (4) fusion codes and (5) supervised local-preserving canonical correlation analysis method. Receiver operating characteristics (ROC), equal error rate, area under the curve of ROC and decidability index (d) are used to show the performance of the system qualitatively and quantitatively.

Soft measurement of wood defects based on LDA feature fusion and compressed sensor images

We proposed a detection method for wood defects based on linear discriminant analysis (LDA) and the use of compressed sensor images. Wood surface images were captured, using a camera Oscar F810C IRF camera, and then the image segmentation was performed, and the defect features were extracted from wood board images. To reduce the processing time, LDA algorithm was used to integrate these features and reduce their dimensions. Features after fusion were used to construct a data dictionary and a compressed sensor was designed to recognize the wood defects types. Of the three major defect types, 50 images live knots, dead knots, and cracks were used to test the effects of this method. The average time for feature fusion and classification was 0.446 ms with the classification accuracy of 94%.