Fisher Discriminant Analysis Research Articles

BoF-SVM-based data intelligence model for detecting dementia

Dementia is a brain condition that impairs the cognitive abilities of an individual. Mild cognitive impairment is a mediator phase of healthy and dementia controls. The motivation of this study is to predict dementia using magnetic resonance imaging data, which is significant for the diagnosis of normal control and dementia patients. The proposed model leverages effective methods like Discrete Wavelet Transform, Bag of Features, and Support Vector Machine. The four wavelets haar, Daubechies, symlets, and coiflets are used for image compression. The results of the proposed data intelligence model are promising in terms of accuracy which is 92.32% which is better than the recently proposed models. Also, the proposed data intelligence model is compared with the models which may use curvelet transform, and shearlet transform and with the methods which have gone without using DWT transforms. The comparisons have also been made with the models that have used other prevalent techniques like Principal Component Analysis, Fisher Discriminant Ratio, and Gray Level Co-occurrence Matrix. The outcomes support the usage of each technique in the suggested data intelligence paradigm.

Molecular quantification for differentiation of fresh and dried Jinqian Baihua She

Freshly-used crude drugs have unique functions and advantages in TCM practice of treating diseases. Jinlong Capsule is a patent traditional Chinese medicine product effective for treatment of hepatocarcinoma, and fresh Jinqian Baihua She (JBS, the body of juvenile Bungarus multicinctus) is one of its important ingredients. The emergence of counterfeit fresh JBS, often identified as dried JBS with almost identical appearance, poses a difficult problem in the quality control of the product. Herein we report a molecular quantification-based method for differentiation of fresh and dried JBS by determining the copy number of a specific DNA marker in the samples. Using species-specific primers and TaqMan probes, we established a real-time quantitative PCR system for amplification of a fragment in the 658-bp cytochrome oxidase subunit I (COI) region from JBS specimens. The amplicon copy number in the muscle tissues ranged from 1.14 × 107 to 4.83 × 107 copies/mg in fresh JBS samples, as compared with 1.13 × 105−8.91 × 106 copies/mg in dried JBS samples. Based upon Fisher discriminant analysis, we used 1.27 × 107 copies/mg as the cut-off value for differentiating fresh and dried JBS, which was validated in the single-blinded validation test of fresh and dried JBS samples. This qPCR system may provide an efficient means for accurate identification of fresh JBS to improve the quality control of the medicinal product.

Unsupervised domain adaptation via transferred local Fisher discriminant analysis

Unsupervised domain adaptation via transferred local Fisher discriminant analysis

RobustDespeckling: Robust speckle noise reduction method using multi-scale and kernel fisher discriminant analysis

In medical ultrasound imaging, speckle noise is an important property since it usually involves worsening the image resolution along with contrast, thus reducing the diagnostic significance of the imaging modality. The "RobustDespeckling" method, which has been proposed in this study to reduce speckle noise, involves extracting the noise-only signal from the image signal through applying the bidimesional empirical mode decomposition (BEMD) method and wavelet transform (WT) in the encapsulated way. Then, wavelet coefficients are thresholded to filter the noisy coefficients by an optimal threshold parameter that is estimated from kernel fisher discriminant analysis (KFDA). Finally, BEMD and WT processes are employed to obtain a de-noised image. The investigation is conducted by way of upsampling and downsampling operations of ultrasound images to analyze the efficiency of different de-noising techniques including the proposed “RobustDespeckling” method. The visual observation and statistical analysis signify that the “RobustDespeckling” method is robust in minimizing the speckle noise of ultrasound images over existing de-noising techniques. Consequently, the method is more competent to preserve the clarity of the edge without losing the significant image detail.

Postmortem Interval Estimation Using Protein Chip Technology Combined with Multivariate Analysis Methods.

To estimate postmortem interval (PMI) by analyzing the protein changes in skeletal muscle tissues with the protein chip technology combined with multivariate analysis methods. Rats were sacrificed for cervical dislocation and placed at 16 ℃. Water-soluble proteins in skeletal muscles were extracted at 10 time points (0 d, 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d and 9 d) after death. Protein expression profile data with relative molecular mass of 14 000-230 000 were obtained. Principal component analysis (PCA) and orthogonal partial least squares (OPLS) were used for data analysis. Fisher discriminant model and back propagation (BP) neural network model were constructed to classify and preliminarily estimate the PMI. In addition, the protein expression profiles data of human skeletal muscles at different time points after death were collected, and the relationship between them and PMI was analyzed by heat map and cluster analysis. The protein peak of rat skeletal muscle changed with PMI. The result of PCA combined with OPLS discriminant analysis showed statistical significance in groups with different time points (P<0.05) except 6 d, 7 d and 8 d after death. By Fisher discriminant analysis, the accuracy of internal cross-validation was 71.4% and the accuracy of external validation was 66.7%. The BP neural network model classification and preliminary estimation results showed the accuracy of internal cross-validation was 98.2%, and the accuracy of external validation was 95.8%. There was a significant difference in protein expression between 4 d and 25 h after death by the cluster analysis of the human skeletal muscle samples. The protein chip technology can quickly, accurately and repeatedly obtain water-soluble protein expression profiles in rats' and human skeletal muscles with the relative molecular mass of 14 000-230 000 at different time points postmortem. The establishment of multiple PMI estimation models based on multivariate analysis can provide a new idea and method for PMI estimation.

Identification of Oak-Barrel and Stainless Steel Tanks with Oak Chips Aged Wines in Ningxia Based on Three-Dimensional Fluorescence Spectroscopy Combined with Chemometrics.

With the increased incidence of wine fraud, a fast and reliable method for wine certification has become a necessary prerequisite for the vigorous development of the global wine industry. In this study, a classification strategy based on three-dimensional fluorescence spectroscopy combined with chemometrics was proposed for oak-barrel and stainless steel tanks with oak chips aged wines. Principal component analysis (PCA), partial least squares analysis (PLS-DA), and Fisher discriminant analysis (FDA) were used to distinguish and evaluate the data matrix of the three-dimensional fluorescence spectra of wines. The results showed that FDA was superior to PCA and PLS-DA in classifying oak-barrel and stainless steel tanks with oak chips aged wines. As a general conclusion, three-dimensional fluorescence spectroscopy can provide valuable fingerprint information for the identification of oak-barrel and stainless steel tanks with oak chips aged wines, while the study will provide some theoretical references and standards for the quality control and quality assessment of oak-barrel aged wines.

Characterization of Animal and Vegetable Oil by Attenuated Total Reflectance – Fourier Transform Infrared (ATR-FTIR) Spectroscopy with Supervised Pattern Recognition and Filter Algorithm

The potential of filter algorithms in improving spectral model performance was evaluated. A total of 329 animal and vegetable oil samples were used to collect attenuated total reflectance – Fourier transform infrared (ATR-FTIR) spectra. Fisher discriminant analysis (FDA), support vector machine (SVM), decision tree (DT), K-nearest neighbor analysis (KNN), and multilayer perceptron neural network (MLP) were considered to build models. Three filter algorithms, finite length unit impulse response filter (FIR), infinite length impulse response filter (IIR) and wavelet transform (WT), were evaluated to enhance the performance of the models. The Morlet basis function was the most suitable wavelet transform, accurately classifying 90.3% of the training set and 94.4% of the test set. The MLP algorithms were demonstrated to be superior to the others. The best performance was obtained using low-pass or band-stop filters that provided 100% accuracy with the MLP model. The reported method is demonstrated to be affordable and easy-to-use in forensic analysis.

Seismic response analysis and distribution prediction of source rocks in a survey of the South China Sea

Identification and prediction of high-quality source rocks is the key to obtaining new resources in the exploration area of Cenozoic basins in offshore China. We research the seismic response and distribution of hydrocarbon source rocks using seismic data, well curves, lithologic interpretation, and geochemical analysis. The target is the source rock development zone of the W Formation in a survey of the South China Sea. The results show that the seismic response of thick layer source rocks differ from surrounding rocks in the seismic profile (strong reflections with opposite polarity at the top and bottom and messy or chaotic reflections inside). Seismic reflections of interlayer source rocks have the characteristics of low frequency and continuous strong amplitude. The dominant frequency and maximum amplitude decrease as the number of mudstone layers increases. Through seismic petrophysical analysis, we have obtained three sensitive parameters of source rock in this survey: clay content, P-wave impedance, and elastic impedance. Different classification methods can realize the classification and prediction of hydrocarbon source rocks, among which the Kernel Fisher discriminant analysis method is the best. The prediction results are consistent with the geological background, geochemical information, and well curves.

Combinatorial Material Strategy: Parallel Synthesis and High-Throughput Screening of WO3 Nanoplates Decorated with Noble Metals for VOCs Sensor

In this study, we report on the rapid preparation of WO3 nanoplates decorated with noble metals and evaluate their gas-sensing performance using a high-throughput screening technique. The incorporation of Pd significantly enhanced the gas-sensing properties, and, among all of the samples tested, the WO3 nanoplate containing 0.3 mol% Pd exhibited the highest response to 100 ppm xylene at 250 °C (Ra/Rg = 131.2), which was almost 56 times greater than that of the pure WO3 sample. Additionally, this sample demonstrated rapid response and recovery times (τresponse = 3.9 s and τrecovery = 189.2 s, respectively). The nanoplate samples were also classified and screened using cluster analysis, and the selected samples were optimized for use in a sensor array. By applying principal component analysis and Fisher discriminant analysis, four typical gases were identified and a potential sensitization mechanism was elucidated.

Research on Glass Classification and Recognition Based on Fisher's Linear Discriminant and Hierarchical Clustering

Buried in the ground, the glass cultural relics weathered, and a large number of internal elements were exchanged. In order to classify the types of glass, this paper first uses Fisher linear discriminant analysis to obtain a binary classification model of glass types, and finds that the prediction accuracy of the model has reached 90%. Considering the possible overfitting of the model, this paper also uses the principal component analysis method to reduce the dimension of the data, thus obtaining the model and determining that its accuracy rate is maintained above 90%. In order to identify the types of unknown glass cultural relics, the multi-category Fisher’s Linear Discriminant method based on principal component analysis was used to identify the subcategories they belonged to.

Component Analysis Model of Glass Relics Based on Logistic Regression and Fisher Linear Discrimination

In recent years, with the attention paid by various countries to the exploitation and protection of cultural relics and the continuous development of industrialization, people have gradually realized the necessity of analyzing the weathering of ancient glass relics. In this paper, the physical characteristics, chemical composition and content of sample cultural relics are analyzed and their types are identified by using multiple algorithms such as Logistic regression model, Fisher linear discrimination, Fisher’s exact test, etc. The study of this issue can provide guiding suggestions for the analysis and classification of ancient cultural materials, and help archaeologists carry out a series of work such as cultural relics protection. The results show that the "type" factor has the greatest relationship with weathering, followed by "color" and "decoration". The regression equation between the sample and weathering chemical composition is established, and it is concluded that "type" and "SnO2" have a large negative correlation with weathering, "MgO" and "SO2" have a large positive correlation, and other variables have a weak correlation. Finally, Fisher linear discriminant method is used to identify the category of unknown glass relics. By estimating the training group and testing the test group, it is found that the algorithm can correctly classify 100.0% of the original grouped cases. We can identify the unknown cultural relics and find that A1, A5, A6 and A7 cultural relics belong to high potassium glass, and the rest belong to lead barium glass. In order to test the accuracy of the model in identifying the category of glass relics, Fisher's exact test algorithm was used. It was found that the classification results were more sensitive to silica in the fluctuation range of - 50% to - 25%, and less sensitive to other chemical components, it is stable.

A classification model for ancient glass artefacts based on Fisher discriminant analysis and K-means clustering

Classification is important for the study of ancient glass objects. Accurate identification and classification is important in presenting the cultural origins, heritage, and changes of glass artefacts. In order to accurately classify ancient glass artefacts, this paper first classifies ancient glass artefacts into two categories based on Fisher discriminant analysis, establishes a K-mean clustering model to subdivide the two types of artefacts, and further conducts sensitivity analysis to establish a classification model for ancient glass artefacts.

Classification and Identification of Ancient Glass Objects based on K-means++ and Fisher Discriminant

The significance of this study is to classify and identify the types of ancient glass products according to their chemical composition. The data were selected from the proportion of chemical compositions that have been analyzed for ancient glass, and the best number of clusters (k) for the division was roughly determined using the elbow rule for the five chemical compositions of high potassium and lead barium, respectively, and brought into the k-means++ algorithm for cluster analysis, and then the final determination of k and the evaluation of the rationality of clustering were performed using the contour coefficient, and finally the Fisher discriminant analysis method based on variable meritocracy combined with eigenvalues, Wilke Lambda, and classification function coefficients to identify unknown categories of glass artifacts. The model used was analyzed and evaluated with good results, and the model is applicable to the classification of ancient glass artifacts and identification of the type to which they belong.

The evaluation of the nutritional quality of common wild edible fungi in Yunnan Province

There are huge resource reserves of wild edible fungi richer in their varieties in Yunnan Province which is located on plateau of low latitude and possesses unique various climate environments and bigger vegetative cover ratios. Moreover, nutrients and flavor substances in the same or various species of wild edible fungi differ greatly with the influence on different components from habitats and geographic areas. So, 5 common wild edible fungi were collected from different areas in Yunnan Province, and several findings were made from this research. Above all, through the evaluation of amino acids, these 5 fungi met the criteria for ideal protein by WHO/FAO, and the nutritional value of protein was ranked as matsutake>truffle>collybia albuminosa>bolete>chanterelle. Next, after the analysis of taste activity values, the ranking of taste was bolete>collybia albuminosa>truffle>matsutake>chanterelle. Subsequently, the ranking of characters was truffle>collybia albuminosa >bolete>matsutake>chanterelle through principal component analysis. Finally, truffle could be completely divided by Fisher discrimination analysis with a bigger difference from others, main in ash, protein, sugar, and polysaccharide, meanwhile, truffle and bolete could be completely divided by orthogonal projections to latent structures discrimination analysis, main in protein, crude fiber, fat, and amino acid. So, there was a more conspicuous difference in nutrients among fungi, through which nutrients combined with multivariate statistics analysis made possible the correct differentiation of small range categories in wild edible fungi, and the correct classification of small range of them could effectively be realized.

Time and Space Complexity Reduction of KFDA-based LTE Modulation Classification

Background: Kernel Fisher discriminant analysis (KFDA) is a nonlinear discrimination technique for improving automatic modulation classification (AMC) accuracy. Our study showed that the higher-order cumulants (HOCs) of the Long-term evolution (LTE) modulation types are nonlinearly separable, so the KFDA technique is a good solution for its modulation classification problem. Still, research papers showed that the KFDA suffers from high time and space computational complexity. Some studies concentrated on reducing the KFDA time complexity while preserving the AMC performance accuracy by finding faster calculation techniques, but unfortunately, they couldn't reduce the space complexity. Objective: This study aims to reduce the time and space computational complexity of the KFDA algorithm while preserving the AMC performance accuracy. Methods: Two new time and space complexity reduction algorithms have been proposed. The first algorithm is the most discriminative dataset points (MDDP) algorithm, while the second is the k-nearest neighbors-based clustering (KNN-C) algorithm Results: The simulation results show that these algorithms could reduce the time and space complexities, but their complexity reduction is a function of signal-to-noise ratio (SNR) values. On the other hand, the KNN-C-based KFDA algorithm has less complexity than the MDDP-based KFDA algorithm Conclusion: The time and space computation complexity of the KFDA could be effectively reduced using MDDP and KNN-C algorithms; as a result, its calculation became much faster and had less storage size.

A thermal comfort evaluation model based on facial skin temperature

This study realized a Fisher discriminant analysis method for thermal comfort evaluation based on facial skin temperature. These models directly determine whether an individual is in a neutral state through neutral facial skin temperature thresholds and have physiological significance. The implementation of the models was based on climate-controlled experimental data. The air temperatures considered included 12, 15, 18, 21, and 24 °C (winter) and 24, 26, 28, 30, and 32 °C (summer). During the experiments, the thermal sensation votes of 40 participants were documented using subjective questionnaires, and facial skin temperatures were continuously measured using high-precision corrected iButtons. The results show that the highest precision of the evaluation model was 80%, and the corresponding false-negative rate (FNR) was 27%. The neutral facial skin temperature ranges were higher during summer than during winter. After 15 min of exposure, the models exhibited higher reliability. The average skin temperature from multiple facial parts was more suitable for evaluating neutral thermal sensations than that from a single facial part. The average skin temperature of the nose, cheeks, and chin is the most appropriate index for future applications of infrared thermal imaging. To a certain extent, the models were further improved by sex. Considering air-conditioning energy conservation, this study provides expanded evaluation models. The recommended average skin temperature exhibited a higher evaluation performance: the highest precision was 86%, and the FNR was 19%. These results provide a theoretical basis for the future use of the facial skin temperature as a feedback signal for air conditioners.

Skill-level classification and performance evaluation for endoscopic sleeve gastroplasty.

We previously developed grading metrics for quantitative performance measurement for simulated endoscopic sleeve gastroplasty (ESG) to create a scalar reference to classify subjects into experts and novices. In this work, we used synthetic data generation and expanded our skill level analysis using machine learning techniques. We used the synthetic data generation algorithm SMOTE to expand and balance our dataset of seven actual simulated ESG procedures using synthetic data. We performed optimization to seek optimum metrics to classify experts and novices by identifying the most critical and distinctive sub-tasks. We used support vector machine (SVM), AdaBoost, K-nearest neighbors (KNN) Kernel Fisher discriminant analysis (KFDA), random forest, and decision tree classifiers to classify surgeons as experts or novices after grading. Furthermore, we used an optimization model to create weights for each task and separate the clusters by maximizing the distance between the expert and novice scores. We split our dataset into a training set of 15 samples and a testing dataset of five samples. We put this dataset through six classifiers, SVM, KFDA, AdaBoost, KNN, random forest, and decision tree, resulting in 0.94, 0.94, 1.00, 1.00, 1.00, and 1.00 accuracy, respectively, for training and 1.00 accuracy for the testing results for SVM and AdaBoost. Our optimization model maximized the distance between the expert and novice groups from 2 to 53.72. This paper shows that feature reduction, in combination with classification algorithms such as SVM and KNN, can be used in tandem to classify endoscopists as experts or novices based on their results recorded using our grading metrics. Furthermore, this work introduces a non-linear constraint optimization to separate the two clusters and find the most important tasks using weights.

Forensic Analysis of Faces on Low-Quality Images using Detection and Recognition Methods

Facial recognition is an essential aspect of conducting criminal action investigations. Captured images from the camera or the recording video can reveal the perpetrator's identity if their faces are deliberately or accidentally captured. However, many of these digital imagery results display the results of image quality that is not good when seen by the human eye. Hence, the facial recognition process becomes more complex and takes longer. This research aims to analyze face recognition on a low-quality image with noise, blur and brightness problem to help digital forensic investigator do an investigation in recognizing faces that the human eye can’t do. The Viola-Jones algorithm method has several processes such as the Haar feature, integral image, adaboost, and cascade classifier for detecting a face in an image. Detected face will be passed to the next process for recognition call Fisher’s Linear Discriminant (FLD), Local Binary Pattern’s (LBP) and Principal Component analysis (PCA). The software's facial recognition feature shows one of the images in the database that the program suspects has the same face as the analyzed face image. In conclusion, from the analysis we determined that LBP approach is the best among the other recognition methods for blur and brightness problem, bet found PCA method is the best for recognize face in noise problem. The software's facial recognition feature shows one of the images in the database that the program suspects has the same face as the analyzed face image. The position of the face object in the image, whether or not there is an additional object that was not previously included in the image in the dataset, as well as the brightness level of an image and the color of the face's skin, all affect the accuracy rates.

Sparse Local Fisher Discriminant Analysis for Gas-Water Two-Phase Flow Status Monitoring With Multisensor Signals

Gas-water two-phase flow has typically stable flow statuses and constantly changing transition flow statuses. Accurate identification and real-time monitoring of flow status are conducive to the in-depth study of two-phase flow and the safe operation of industrial process. A monitoring strategy based on sparse local Fisher discriminant analysis (SLFDA) is proposed in this article. First, multisensor signals are obtained to reflect flow process information. Second, the least absolute shrinkage and selection operator is used to find the sparse discriminant directions to determine the key variables relevant to the flow process from multiple sensor signals. Then, the weight coefficient matrixes of SLFDA keep the original structure of the same flow status data and make the data of different flow statuses more separated, which distinguish different flow statuses to the maximum extent. Finally, two monitoring indexes including the discriminant index and the stability index are established to analyze the dynamic flow process, which enable a concurrent monitoring of both flow evolution and instability to realize fine-scale description of flow process. SLFDA can monitor various flow statuses through only one projection discriminant matrix by transforming high-dimensional signals into features representing the flow characteristics, which avoids model traversal and improves monitoring efficiency. Further study of flow status features provides meaningful physical interpretation and in-depth process analysis with consideration of actual flow process. The application on the data of gas-water two-phase flow in horizontal pipe demonstrates the feasibility and efficacy of the method.

Quantitative assessment of adulteration of coconut oil using transmittance multispectral imaging.

Economical to a fault, coconut oil is a commodity related to fraudulent activities such as oil adulteration for undue profits. Unfortunately, the conventional methods used in the detection of adulteration and toxicants are laborious, destructive, and time-consuming. Hence, it is imperative to engineer a non-destructive and rapid screening test with sufficient accuracy. To that end, the proposed work has an in-house developed imaging system hardware and a method to estimate relevant quality parameters from multispectral imagery. Multispectral images of adulterated coconut oil were analyzed through a cascade of statistical algorithms: Fisher Discriminant Analysis and Bhattacharyya distance respectively. In this work, a functional relationship was developed for the estimation of adulteration level that recorded an R2 of 0.9876 for the training samples and an MSE of 0.0029 for the testing samples. Besides, the proposed imaging system offers flexibility on post-processing of raw measurements as the algorithm is designed to operate from raw multispectral images. In addition, the developed imaging system is economical in its capacity to estimate the adulteration of coconut oil with remarkable accuracy considering the low cost of production. Moreover, the proposed work validates the use of multispectral imagery as an initial screening technique instead of expensive spectroscopy methods.