Subspace Linear Discriminant Analysis Research Articles

Investigating the Performance of Wearable Motion Sensors on recognizing falls and daily activities via machine learning

With sensor-based wearable technologies, high precision monitoring and recognition of human physical activities in real time is becoming more critical to support the daily living requirements of the elderly. The use of sensor technologies, including accelerometers (A), gyroscopes (G), and magnetometers (M) is mostly encountered in work focused on assistive technology, ambient intelligence, context-aware systems, gait and motion analysis, sports science, and fall detection. The classification performance of four sensor type combinations is investigated through the use of four machine learning algorithms: support vector machines (SVMs), Manhattan k-nearest neighbor classifier (M.k-NN), subspace linear discriminant analysis (SLDA), and ensemble bagged decision tree (EBDT). In this context, a large dataset containing 2520 tests performed by 14 volunteers containing 16 activities of daily living (ADLs) and 20 falls was employed. In binary (fall vs. ADL) and multi-class activity (36 activities) recognition, the highest classification accuracy rate was obtained by the SVM (99.96%) and M.k-NN (95.27%) classifiers, respectively, with the AM sensor type combination in both cases. We also made our dataset publicly available to lay the groundwork for new research.

Hyperspectral reflectance for non-invasive early detection of black shank disease in flue-cured tobacco

Flue-cured tobacco (Nicotiana tabacum L.) is a high value-per-acre crop that is intensively managed to optimise the yield of high-quality cured leaf. A 15-day study assessed the potential of hyperspectral reflectance data for detecting Phytophthora nicotianae (black shank) incidence in flue-cured tobacco. Hyperspectral reflectance data were taken from a commercial flue-cured tobacco field with a progressing black shank infestation. The effort encompassed two key objectives. First, develop hyperspectral indices and/or machine learning classification models capable of detecting Phytophthora nicotianae (black shank) incidence in flue-cured tobacco. Second, evaluate the model’s ability to separate pre-symptomatic plants from healthy plants. Two hyperspectral indices were developed to detect black shank incidence based on differences in the spectral profiles of asymptomatic flue-cured tobacco plants compared to those with black shank symptoms. While one of the indices is a broad-band index and the other uses narrow wavelength values, the statistical difference between the two indices was not significant and both provided an accurate classification of symptomatic plants. Further analysis of the indices showed significant differences between the index values of healthy and symptomatic plants (α = 0.05). In addition, the indices were able to detect black shank symptoms pre-symptomatically (α = 0.09). Subspace linear discriminant analysis, a machine learning classification, was also used for prediction of black shank incidence with up to 85.7% classification accuracy. The implications of using either spectral indices or machine learning for classification for future black shank research are discussed.

A perinatal stress calculator for the neonatal intensive care unit: an unobtrusive approach

Objective: Early experience of pain and stress in the neonatal intensive care unit is known to have an effect on the neurodevelopment of the infant. However, an automated method to quantify the procedural pain or perinatal stress in premature patients does not exist. Approach: In the current study, EEG and ECG data were collected for more than 3 hours from 136 patients in order to quantify stress exposure. Specifically, features extracted from the EEG and heart-rate variability in both quiet and non-quiet sleep segments were used to develop a subspace linear-discriminant analysis stress classifier. Main results: The main novelty of the study lies in the absence of intrusive methods or pain elicitation protocols to develop the stress classifier. Three main findings can be reported. First, we developed different stress classifiers for the different age groups and stress intensities, obtaining an area under the curve in the range [0.78-0.93] for non-quiet sleep and [0.77-0.96] for quiet sleep. Second, a dysmature EEG was found in patients under stress. Third, an enhanced cortical connectivity and increased brain–heart communication was correlated with a higher stress load, while the autonomic activity did not seem to be associated to stress exposure. Significance: The results shed a light on the pain and stress processing in preterm neonates, suggesting that software tools to investigate dysmature EEG might be helpful to assess stress load in premature patients. These results could be the foundation to assess the impact of stress on infants’ development and to tune preventive care.

A Bradycardia-Based Stress Calculator for the Neonatal Intensive Care Unit: A Multisystem Approach.

Early life stress in the neonatal intensive care unit (NICU) can predispose premature infants to adverse health outcomes and neurodevelopment delays. Hands-on-care and procedural pain might induce apneas, hypoxic events, and sleep-wake disturbances, which can ultimately impact maturation, but a data-driven method based on physiological fingerprints to quantify early-life stress does not exist. This study aims to provide an automatic stress detector by investigating the relationship between bradycardias, hypoxic events and perinatal stress in NICU patients. EEG, ECG, and SpO2 were recorded from 136 patients for at least 3 h in three different monitoring groups. In these subjects, the stress burden was assessed using the Leuven Pain Scale. Different subspace linear discriminant analysis models were designed to detect the presence or the absence of stress based on information in each bradycardic spell. The classification shows an area under the curve in the range [0.80–0.96] and a kappa score in the range [0.41–0.80]. The results suggest that stress seems to increase SpO2 desaturations and EEG regularity as well as the interaction between the cardiovascular and neurological system. It might be possible that stress load enhances the reaction to respiratory abnormalities, which could ultimately impact the neurological and behavioral development.

Spectral feature extraction of hyperspectral remote sensing images based on class pair-weighted criterion

Hyperspectral remote sensing images (HRSIs) face the problem of extracting efficient discriminant features when foreign objects with similar spectra and outlier classes exist. For solving the problem, first, we derive a class pair (CP) form of the classical Fisher criterion, namely CP Fisher criterion, where the between-class scatter matrix and within-class scatter matrix are both decomposed into the form of CPs. And then, a CP-weighted criterion is proposed to weight the between-class scatter matrix and within-class scatter matrix of each CP according to its separability so that the separabilities of CPs are balanced in the CP-weighted feature subspace. The Bayesian classifier and k-nearest neighbors classifier are used to evaluate the feature extraction performance. Experimental results on three real HRSIs show that the presented CP-weighted subspace method improves the classification accuracies of the CPs with small separabilities so that it is superior to the linear discriminant analysis subspace method and the weighted pairwise Fisher subspace method.

Cross domain association using transfer subspace learning

Transfer learning has gained more attention recently by utilizing knowledge acquired from one domain to advance a learning performance in another domain. Existing homogeneous transfer learning methods have progressed to a point where feature spaces are common in training and testing domains. However, heterogeneous transfer learning is still in its nascent stage where features of training and testing domains are different. Taking this into account, Bregman Divergence Regularization is used to minimize the probability distribution difference between training and testing domains and to take them together to a shared subspace. To discriminate data within individual domains, a projection matrix is obtained using Fisher Linear Discriminant Analysis subspace learning algorithm. Experimentation is performed on two efficiently used biometrics: the face and fingerprint. Two types of cross-domain settings are used: (1) Face + Finger2Finger where training samples come from face (labeled samples) and fingerprint (unlabeled samples) data sets, while testing is performed on a fingerprint dataset. (2) Finger + Face2Face where training samples come from fingerprint (labeled samples) and face (unlabeled samples) data sets while testing is performed on a face dataset. This paper proposes a cross domain association between face and fingerprint that finds utility in forensic applications.

A novel discriminant multiscale representation for ear recognition

This paper proposes a novel representation for ear recognition. It introduces a new alternative of binarised statistical image features based on multiscale framework. The proposed representation allows capturing the image content at multiple resolutions. The recognition accuracy can be enhanced by the following steps. First, for a given ear image, a set of multiscale response images are derived from the bank of binarised statistical image features (B-BSIF) filter. Second, the obtained response images are summarised by concatenating their histograms, which are obtained at each scale. Finally, a discriminative ear image representation is build by projecting the above mentioned histograms into a linear discriminant analysis subspace. The proposed representation is applied on three public databases: IIT Delhi-1, IIT Delhi-2 and USTB. The obtained recognition accuracy confirms its performance than the recent existing methods.

Big Feature Data Analytics: Split and Combine Linear Discriminant Analysis (SC-LDA) for Integration Towards Decision Making Analytics

This paper introduces a novel big feature data analytics scheme for integration toward data analytics with decision making. In this scheme, a split and combine approach for a linear discriminant analysis (LDA) algorithm termed SC-LDA is proposed. The SC-LDA replaces the full eigenvector decomposition of LDA with much cheaper eigenvector decompositions on smaller sub-matrices, and then recombines the intermediate results to obtain the exact reconstruction as for the original algorithm. The splitting or decomposition can be further applied recursively to obtain a multi-stage SC-LDA algorithm. The smaller sub-matrices can then be computed in parallel to reduce the time complexity for big data applications. The approach is discussed for an LDA algorithm variation (LDA/QR), which is suitable for the analytics of Big Feature data sets. The projected data vectors into the LDA subspace can then be integrated toward the decision-making process involving classification. Experiments are conducted on real-world data sets to confirm that our approach allows the LDA problem to be divided into the size-reduced sub-problems and can be solved in parallel while giving an exact reconstruction as for the original LDA/QR.

Makeup-invariant face identification and verification using fisher linear discriminant analysis-based Gabor filter bank and histogram of oriented gradients

Non-permanent facial makeup is one of the most difficult problems inhibiting face recognition systems in security applications. In this paper, a new method is proposed for makeup-invariant face identification and verification. Face images from the virtual makeup (VMU) and YouTubemakeup (YMU) datasets were subjected to the Gabor filtering and histogram of oriented gradients (HOG) methods for feature extraction. The Gabor and HOG features were concatenated to generate the final feature vectors and subsequently reduced using the fisher linear discriminant analysis subspace. The reduced features were classified using the city block distance (CBD), Euclidean distance (EUC), cosine similarity measure (CSM) and whitened cosine similarity measure (WCSM). The CSM achieved the best recognition rates out of the four metrics used. Performance evaluation of these metrics produced identification and verification rates of 100% and 100% for the VMU database, and 72.52% and 79.47% for the YMU database, respectively. The developed method outperformed several state-of-the-art methods initially exploited.

Edge-enhanced bi-dimensional empirical mode decomposition-based emotion recognition using fusion of feature set

Emotion recognition has been of great interest in psychology, machine intelligence, human–machine interaction and biomedical fields. This paper proposes a novel soft computing technique for facial emotion recognition by introducing edge- enhanced bidimensional empirical mode decomposition (EEBEMD) as a feature extraction tool for facial emotion recognition. Facial images are subjected to optimized cost function-based self-guided edge enhancement algorithm. BEMD has been applied on the edge- enhanced facial images, and the first four intrinsic mode functions (IMFs) and the residue have been calculated. On the basis of an empirical analysis, the first IMF is selected for further analysis. A proposed fusion model that consists of selected features from the gray-level co-occurrence matrix, the histogram of oriented gradients and the local ternary pattern of the IMF response is fed to a recursive feature elimination-based algorithm to select the appropriate feature subsets for classification. These feature vectors have been trained in three machine learning algorithms namely multi-class SVM, ELM with RBF kernel and k-NN classifier independently. The IMFs have been subjected to principal component analysis and linear discriminant analysis (LDA) algorithm successively for dimensionality reduction, and the facial images with different emotions have been clustered in different zones in the LDA subspace. The proposed method demonstrates promising accuracy when tested on the JAFFE database, Cohn–Kanade database and the eNTERFACE database.

An ensemble of patch-based subspaces for makeup-robust face recognition

Recent research has demonstrated the negative impact of makeup on automated face recognition. In this work, we introduce a patch-based ensemble learning method, which uses multiple subspaces generated by sampling patches from before-makeup and after-makeup face images, to address this problem. In the proposed scheme, each face image is tessellated into patches and each patch is represented by a set of feature descriptors, viz., Local Gradient Gabor Pattern (LGGP), Histogram of Gabor Ordinal Ratio Measures (HGORM) and Densely Sampled Local Binary Pattern (DS-LBP). Then, an improved Random Subspace Linear Discriminant Analysis (SRS-LDA) method is used to perform ensemble learning by sampling patches and constructing multiple common subspaces between before-makeup and after-makeup facial images. Finally, Collaborative-based and Sparse-based Representation Classifiers are used to compare feature vectors in this subspace and the resulting scores are combined via the sum-rule. The proposed face matching algorithm is evaluated on the YMU makeup dataset and is shown to achieve very good results. It outperforms other methods designed specifically for the makeup problem.

Noise-Adjusted Subspace Discriminant Analysis for Hyperspectral Imagery Classification

Linear discriminant analysis (LDA) is a popular approach for dimensionality reduction for pattern classification; however, its performance is often degraded when samples are too few, particularly when the dimensionality of the input feature space is excessively high. The classic solution to the small-sample-size problem is to implement LDA in a principal component (PC) subspace, i.e., a strategy known as subspace LDA. This latter approach is extended by coupling LDA and noise-adjusted HSI analysis in order to provide noise-robust feature extraction and classification of high-dimensional data. An extension of the proposed approach in a kernel-induced space is also studied. The resulting noise-adjusted subspace discriminant analysis is evaluated using hyperspectral imagery, with experimental results demonstrating that the proposed approach provides not only superior classification performance, as compared with traditional methods, but also effective dimensionality reduction for classification even in the presence of noise.

An improvement of incremental recursive fisher linear discriminant for online feature extraction

This paper proposes a new online feature extraction method called the Incremental Recursive Fisher Linear Discriminant (IRFLD), whose batch learning algorithm, referred to as RFLD, was proposed by Xiang and colleagues. In the conventional Linear Discriminant Analysis (LDA), the number of discriminant vectors is limited to the number of classes minus one due to the rank of the between-class covariance matrix. However, RFLD and the proposed IRFLD can break this limit; that is, an arbitrary number of discriminant vectors can be obtained. In the proposed IRFLD, the Incremental Linear Discriminant Analysis (ILDA) of Pang and colleagues is extended in such a way that effective discriminant vectors are recursively searched for in the complementary space of a conventional discriminant subspace. In addition, to estimate a suitable number of effective discriminant vectors, the classification accuracy is evaluated using the cross-validation method in an online manner. For this purpose, validation data are obtained by performing k-means clustering on incoming training data and the previous validation data. The performance of IRFLD is evaluated for 16 benchmark data sets. The experimental results show that the final classification accuracies of IRFLD are always better than those of ILDA. We also show that this performance improvement is attained by adding discriminant vectors in a complementary LDA subspace. © 2013 Wiley Periodicals, Inc. Electron Comm Jpn, 96(4): 29–40, 2013; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/ecj.10430

Fusion of face and iris biometrics using local and global feature extraction methods

Fusion of multiple biometrics combines the strengths of unimodal biometrics to achieve improved recognition accuracy. In this study, face and iris biometrics are used to obtain a robust recognition system by using several feature extractors, score normalization and fusion techniques. Global and local feature extractors are used to extract face and iris features separately, and then, the fusion of these modalities is performed on different subsets of face and iris image databases of ORL, FERET, CASIA and UBIRIS. The proposed method uses Local Binary Patterns local feature extractor and subspace Linear Discriminant Analysis global feature extractor on face and iris images, respectively. Face and iris scores are normalized using tanh normalization, and then, Weighted Sum Rule is applied for the fusion of these two modalities. Improved recognition accuracies are achieved compared to the individual systems and multimodal systems using other local or global feature extractors for both modalities.

Image Recognition using Coefficient of Correlation and Structural SIMilarity Index in Uncontrolled Environment

The goal of this paper is to analyse and improve the performance of metrics like Coefficient of Correlation (CoC) and Structural Similarity Index (SSIM) for image recognition in real-time environment. The main novelties of the methods are; it can work under uncontrolled environment and no need to store multiple copies of the same image at different orientations. The values of CoC and SSIM get changed if images are rotated or flipped or captured under bad/highly illuminated conditions. To increase the recognition accuracy, the input test image is pre-processed. First, discrete wavelet transform is applied to recognize the image captured under bad illuminated and dull lightning conditions. Second, to make the method rotation invariant, the test image is compared against the stored database image without and with rotations in the horizontal, vertical, diagonal, reverse diagonal and flipped directions. The image recognition performance is evaluated using the Recognition Rate and Rejection Rate. The results indicate that recognition performance of Correlation Coefficient and SSIM gets improved with rotations and discrete wavelet transform. Also it was observed that CoC with proposed modifications yield better results as compared to state of the art enhanced Principal Component Analysis and Enhanced Subspace Linear Discriminant Analysis. Keywords—Image Recognition, Discrete Wavelet Transforms, Correlation Coefficient, Structural Similarity Index Metrics.

An Improvement of Incremental Recursive Fisher Linear Discriminant for Online Feature Extraction

This paper proposes a new online feature extraction method called Incremental Recursive Fisher Linear Discriminant (IRFLD) whose batch learning algorithm called RFLD has been proposed by Xiang et al. In the conventional Linear Discriminant Analysis (LDA), the number of discriminant vectors is limited to the number of classes minus one due to the rank of the between-class covariance matrix. However, RFLD and the proposed IRFLD can break this limit; that is, an arbitrary number of discriminant vectors can be obtained. In the proposed IRFLD, the Pang et al.'s Incremental Linear Discriminant Analysis (ILDA) is extended such that effective discriminant vectors are recursively searched for the complementary space of a conventional discriminant subspace. In addition, to estimate a suitable number of effective discriminant vectors, the classification accuracy is evaluated with a cross-validation method in an online manner. For this purpose, validation data are obtained by performing the k-means clustering against incoming training data and previous validation data. The performance of IRFLD is evaluated for 16 benchmark data sets. The experimental results show that the final classification accuracies of IRFLD are always better than those of ILDA. We also reveal that this performance improvement is attained by adding discriminant vectors in a complementary LDA subspace.

WAVELET-FACE BASED SUBSPACE LDA METHOD TO SOLVE SMALL SAMPLE SIZE PROBLEM IN FACE RECOGNITION

Linear Discriminant Analysis (LDA) is a popular statistical method for both feature extraction and dimensionality reduction in face recognition. The major drawback of LDA is the so-called small sample size (3S) problem. This problem always occurs when the total number of training samples is smaller than the dimension of feature space. Under this situation, the within-class scatter matrix Sw becomes singular and LDA approach cannot be implemented directly. To overcome the 3S problem, this paper proposes a novel wavelet-face based subspace LDA algorithm. Wavelet-face feature extraction and dimensionality reduction are based on two-level D4-filter wavelet transform and discarding the null space of total class scatter matrix St. It is shown that our obtained projection matrix satisfies the uncorrelated constraint conditions. Hence in the sense of statistical uncorrelation, this projection matrix is optimal. The proposed method for face recognition has been evaluated with two public available databases, namely ORL and FERET databases. Comparing with existing LDA-based methods to solve the 3S problem, our method gives the best performance.

Limitations of Principal Components Analysis for Hyperspectral Target Recognition

Dimensionality reduction is a necessity in most hyperspectral imaging applications. Tradeoffs exist between unsupervised statistical methods, which are typically based on principal components analysis (PCA), and supervised ones, which are often based on Fisher's linear discriminant analysis (LDA), and proponents for each approach exist in the remote sensing community. Recently, a combined approach known as subspace LDA has been proposed, where PCA is employed to recondition ill-posed LDA formulations. The key idea behind this approach is to use a PCA transformation as a preprocessor to discard the null space of rank-deficient scatter matrices, so that LDA can be applied on this reconditioned space. Thus, in theory, the subspace LDA technique benefits from the advantages of both methods. In this letter, we present a theoretical analysis of the effects (often ill effects) of PCA on the discrimination power of the projected subspace. The theoretical analysis is presented from a general pattern classification perspective for two possible scenarios: (1) when PCA is used as a simple dimensionality reduction tool and (2) when it is used to recondition an ill-posed LDA formulation. We also provide experimental evidence of the ineffectiveness of both scenarios for hyperspectral target recognition applications.

Choosing Parameters of Kernel Subspace LDA for Recognition of Face Images Under Pose and Illumination Variations

This paper addresses the problem of automatically tuning multiple kernel parameters for the kernel-based linear discriminant analysis (LDA) method. The kernel approach has been proposed to solve face recognition problems under complex distribution by mapping the input space to a high-dimensional feature space. Some recognition algorithms such as the kernel principal components analysis, kernel Fisher discriminant, generalized discriminant analysis, and kernel direct LDA have been developed in the last five years. The experimental results show that the kernel-based method is a good and feasible approach to tackle the pose and illumination variations. One of the crucial factors in the kernel approach is the selection of kernel parameters, which highly affects the generalization capability and stability of the kernel-based learning methods. In view of this, we propose an eigenvalue-stability-bounded margin maximization (ESBMM) algorithm to automatically tune the multiple parameters of the Gaussian radial basis function kernel for the kernel subspace LDA (KSLDA) method, which is developed based on our previously developed subspace LDA method. The ESBMM algorithm improves the generalization capability of the kernel-based LDA method by maximizing the margin maximization criterion while maintaining the eigenvalue stability of the kernel-based LDA method. An in-depth investigation on the generalization performance on pose and illumination dimensions is performed using the YaleB and CMU PIE databases. The FERET database is also used for benchmark evaluation. Compared with the existing PCA-based and LDA-based methods, our proposed KSLDA method, with the ESBMM kernel parameter estimation algorithm, gives superior performance.

Human Face Image Searching System Using Sketches

This paper reports a human face image searching system using sketches. A two-phase method, namely, sketch-to-mug-shot matching and human face image searching using relevance feedback, is designed and developed. In the sketch-to-mug-shot matching phase, we have developed a facial feature matching algorithm using local and global features. A point distribution model is employed to represent local facial features while the global feature consists of a set of the geometrical relationship between facial features. It is found that the performance of the sketch-to-mug-shot matching is good if the sketch image looks like the mug shot image in the database. However, in some situations, it is hard to construct a sketch that looks like the photograph. To overcome this limitation, this paper makes use of the concept of ldquohuman-in-the-looprdquo and proposes a human face image searching algorithm using relevance feedback in the second phase. Positive and negative samples will be collected from the user. A feedback algorithm that employs subspace linear discriminant analysis for online learning of the optimal projection for face representation is then designed and developed. The proposed system has been evaluated using the FERET database and a Japanese database with hundreds of individuals. The results are encouraging.