Nonlinear Feature Extraction Technique Research Articles

A nonlinear feature extraction approach for speech emotion recognition using VMD and TKEO

Speech emotion recognition (SER) is still a challenging research area in human–computer interaction-based systems. This paper proposed a nonlinear feature extraction technique to improve the classification performance of the SER system. The proposed method explores variational mode decomposition (VMD) with the Teager-Kaiser energy operator (TKEO) for the SER. First, VMD decomposes a speech signal into modes, and then the nonlinear TKEO operator is applied to each mode to obtain a time series. The VMD-TKEO preprocessed signal is used to extract the global features based on Energy, Pitch frequency and Mel frequency cepstral coefficients. The features are statistically examined using the Kruskal-Wallis test. The resultant feature set is examined over the support vector machine and its variants for emotion classification. The Ryerson Audio-Visual database is used for the experiment, and different emotion classification problems are formulated. Finally, the accuracy of the proposed SER architecture is quantitatively analyzed, which outperforms the other existing architectures.

Exploring Dimensionality Reduction Techniques in Multilingual Transformers

In scientific literature and industry, semantic and context-aware Natural Language Processing-based solutions have been gaining importance in recent years. The possibilities and performance shown by these models when dealing with complex Human Language Understanding tasks are unquestionable, from conversational agents to the fight against disinformation in social networks. In addition, considerable attention is also being paid to developing multilingual models to tackle the language bottleneck. An increase in size has accompanied the growing need to provide more complex models implementing all these features without being conservative in the number of dimensions required. This paper aims to provide a comprehensive account of the impact of a wide variety of dimensional reduction techniques on the performance of different state-of-the-art multilingual siamese transformers, including unsupervised dimensional reduction techniques such as linear and nonlinear feature extraction, feature selection, and manifold techniques. In order to evaluate the effects of these techniques, we considered the multilingual extended version of Semantic Textual Similarity Benchmark (mSTSb) and two different baseline approaches, one using the embeddings from the pre-trained version of five models and another using their fine-tuned STS version. The results evidence that it is possible to achieve an average reduction of 91.58% pm 2.59% in the number of dimensions of embeddings from pre-trained models requiring a fitting time 96.68% pm 0.68% faster than the fine-tuning process. Besides, we achieve 54.65% pm 32.20% dimensionality reduction in embeddings from fine-tuned models. The results of this study will significantly contribute to the understanding of how different tuning approaches affect performance on semantic-aware tasks and how dimensional reduction techniques deal with the high-dimensional embeddings computed for the STS task and their potential for other highly demanding NLP tasks.

Feature Extraction based on Empirical Mode Decomposition for Shapes Recognition of Buried Objects by Ground Penetrating Radar

Ground penetrating radar (GPR) is one of the promising non-destructive imaging tools investigations for shallow subsurface exploration such as locating and mapping the buried utilities. In practical applications, GPR images could be noisy due to the system noise, the heterogeneity of the medium, and mutual wave interactions thus, it is a complex task to recognizing the hyperbolic signature of buried objects from GPR images. Therefore, this paper aims to develop nonlinear feature extraction technique of using Empirical Mode Decomposition (EMD) in recognizing the four geometrical shapes (cubic, cylindrical, disc and spherical) from GPR images. A pre-processing step of isolating hyperbolic signature from different background was first employed by mean of Region of Interest (ROI). The hyperbolic signature that describes the shapes was extracted using EMD decomposition to obtain a set of significant features. In this framework, the hyperbolic pattern was decomposed of using EMD, to produce a small set of intrinsic mode functions (IMF) via sifting process. The IMF properties of the signature that exhibit the unique pattern was used as potential features to differentiate the geometrical shapes of buried objects. The extracted IMF features were then fed into machine learning classifier namely Support Vector Machines. To evaluate the effectiveness of the proposed method, a set data collection of GPR-images has been acquired. The experimental results show that the recognition rate of using IMF features was achieved 99.12% accuracy in recognizing the shapes of buried objects whose shows the promising result.

Performance Evaluation of Kernel-Based Feature Extraction Techniques for Face Recognition System

Face recognition is considered to be one of the most reliable biometrics where security issues are of concerned. Feature extraction which is a functional block of a face recognition system becomes a critical problem when there is need to obtain the best feature with minimum classification error and low running time. Most existing face recognition systems have adopted different non-linear feature extraction techniques for face recognition but identification of the most suitable non-linear kernel variants for these systems remain an open problem. Hence, this research work analyzed the performance of three kernel feature extraction technique (Kernel Principal Component Analysis, Kernel Linear Discriminant Analysis and Kernel Independent Component Analysis) for face recognition system. A database of 360 face images was created by obtaining facial images from LAUTECH Biometric Research Group consisting of six facial expressions of 60 persons. Images were preprocessed (gray scaling, cropping and histogram equalization) and the kernel variants were used to extract distinctive features and reduce the dimensionality of each of the images from 600x800 pixels to four smaller dimensions: 50x50, 100x100, 150x150 and 200x200 pixels. Euclidean Distance similarity measure was used for classification. The performance of the three kernel variants was evaluated for face recognition system using 180 images for training and 180 images for testing using the following metrics: Recognition Accuracy (RA) and Recognition Time (RT). Empirical result indicate that KLDA performs best for face recognition system with an average accuracy of 94.52%. The larger image dimension also results in better recognition performance. We intend to experiment on other classifiers for face recognition system in our future work. Keywords— Biometrics, Face, Feature extraction, Kernel, KICA, KPCA, KLDA, Linear, Non-linear;

A nonlinear Kernel-based adaptive learning-by-examples method for robust NDT/NDE of conductive tubes

ABSTRACTIn this work, the real-time non-destructive testing and evaluation (NDT/NDE) of faulty conductive tubes from eddy current (EC) measurements is addressed and solved in a computationally efficient way by means of an innovative learning-by-examples (LBE) methodology. More specifically, the estimation of the descriptors of a defect embedded within the cylindrical structure under test (SUT) is yielded by combining a non-linear feature extraction technique with an adaptive sampling strategy able to uniformly explore the arising feature space. Predictions are then performed during the on-line testing phase by means of a support vector regression (SVR). Representative results from a numerical/experimental validation are reported to assess the effectiveness of the proposed approach also in comparison with competitive state-of-the-art approaches.

A Fixed-Point Online Kernel Principal Component Extraction Algorithm

Kernel principal component analysis (KPCA) is a powerful and widely applied nonlinear feature extraction technique. However, as originally proposed, KPCA may be cumbersome or infeasible in large-scale datasets, which motivated the development of low-complexity iterative extraction algorithms, mainly aiming image processing applications. Recently, some online KPCA extraction algorithms were proposed, but most of them suffer from low-convergence speed. This paper proposes a new algorithm based on fixed-point iterative equations for KPCA extraction, expanding kernel components using a compact dictionary, dynamically built from data, according to a user-defined accuracy parameter. The algorithm relies on simple equations, can track nonstationary environments, and requires reduced storage, enabling its use in real-time applications operating in low-cost embedded hardware. Results involving open-access image datasets show improved accuracy and convergence speed, as well as permitted effective improvements in practical image applications, as compared to state-of-art online KPCA techniques.

Reconstruction of security region with high information content using effective attributes extraction

Summary Security region-based power system operation guides operators in a fast and relevant decision-making process as to what conditions require operators' actions. The key to this approach is the manner in which effective attributes are extracted among the network's operating variables. This paper reconstructs the security region of power system with high information content. The security region of power system is constructed by means of the static security boundaries such as static voltage stability limit, generator output power limit, and transformer and transmission line thermal limits by increasing loading factor along various directions until these limits are violated. Then, the security region is described by means of effective features set that contain maximum information content. In this paper, nonlinear feature extraction techniques have been proposed to dimensionality reduction and relevant extraction of power system operating variables. Convex and nonconvex techniques of feature extraction are analyzed based on a mutual information index and generalization error. The simulation results on the IEEE 9-bus and the IEEE 118-bus test systems demonstrate the effectiveness of the proposed method for describing the security region by considering high information content.

Voice assessments for detecting patients with Parkinson’s diseases using PCA and NPCA

In this study, we wanted to discriminate between two groups of people. The database used in this study contains 20 patients with Parkinson's disease and 20 healthy people. Three types of sustained vowels (/a/, /o/ and /u/) were recorded from each participant and then the analyses were done on these voice samples. Firstly, an initial feature vector extracted from time, frequency and cepstral domains. Then we used linear and nonlinear feature extraction techniques, principal component analysis (PCA), and nonlinear PCA. These techniques reduce the number of parameters and choose the most effective acoustic features used for classification. Support vector machine with its different kernel was used for classification. We obtained an accuracy up to 87.50 % for discrimination between PD patients and healthy people.

Expert system based on computer vision to estimate the content of impurities in olive oil samples

The determination of the content of impurities is a very frequent analysis performed on virgin olive oil samples, but the official method is quite work-intensive, and it would be convenient to have an alternative approximate method to evaluate the performance of the impurity removal process. In this work we develop a system based on computer vision and pattern recognition to classify the content of impurities of the olive oil samples in three sets, indicative of the goodness of the separation process of olive oil after its extraction from the paste. Starting from the histograms of the channels of the Red–Green–Blue (RGB), CIELAB and Hue-Saturation-Value (HSV) color spaces, we construct an initial input parameter vector and perform a feature extraction previous to the classification. Several linear and non-linear feature extraction techniques were evaluated, and the classifiers used were Support Vector Machines (SVMs) and Artificial Neural Networks (ANNs). The best classification rate achieved was 87.66%, obtained using Kernel Principal Components Analysis (KPCA) and a grade-3-polynomial kernel SVM. The best result using ANNs was 82.38%, yielded by the use of Principal Component Analysis (PCA) with the Perceptron.

Blind vibration component separation and nonlinear feature extraction applied to the nonstationary vibration signals for the gearbox multi-fault diagnosis

Fault diagnosis of gearboxes, especially the gears and bearings, is of great importance to the long-term safe operation. An unexpected damage on the gearbox may break the whole transmission line down. It is therefore crucial for engineers and researchers to monitor the health condition of the gearbox in a timely manner to eliminate the impending faults. However, useful fault detection information is often submerged in heavy background noise. Thereby, a new fault detection method for gearboxes using the blind source separation (BSS) and nonlinear feature extraction techniques is presented in this paper. The nonstationary vibration signals were analyzed to reveal the operation state of the gearbox. The kernel independent component analysis (KICA) algorithm was used hereby as the BSS approach for the mixed observation signals of the gearbox vibration to discover the characteristic vibration source associated with the gearbox faults. Then the wavelet packet transform (WPT) and empirical mode decomposition (EMD) nonlinear analysis methods were employed to deal with the nonstationary vibrations to extract the original fault feature vector. Moreover, the locally linear embedding (LLE) algorithm was performed as the nonlinear feature reduction technique to attain distinct features from the feature vector. Lastly, the fuzzy k-nearest neighbor (FKNN) was applied to the fault pattern identification of the gearbox. Two case studies were carried out to evaluate the effectiveness of the proposed diagnostic approach. One is for the gear fault diagnosis, and the other is to diagnose the rolling bearing faults of the gearbox. The nonstationary vibration data was acquired from the gear and rolling bearing fault test-beds, respectively. The experimental test results show that sensitive fault features can be extracted after the KICA processing, and the proposed diagnostic system is effective for the multi-fault diagnosis of the gears and rolling bearings. In addition, the proposed method can achieve higher performance than that without KICA processing with respect to the classification rate.

Dynamic neuro-fuzzy local modeling system with a nonlinear feature extraction for the online adaptive warning system of river temperature affected by waste cooling water discharge

This paper proposes a dynamic modeling methodology based on a dynamic neuro-fuzzy local modeling system (DNFLMS) with a nonlinear feature extraction technique for an online dynamic modeling task. Prior to model building, a nonlinear feature extraction technique called the Gamma test (GT) is proposed to compute the lowest mean squared error (MSE) that can be achieved and the quantity of data required to obtain a reliable model. Two different DNFLMS modes are developed: (1) an online one-pass clustering and the extended Kalman filtering algorithm (mode 1); and (2) hybrid learning algorithm (mode 2) of extended Kalman filtering algorithm with a back-propagation algorithm trained to the estimated MSE and number of data points determined by a nonlinear feature extraction technique. The proposed modeling methodology is applied to develop an online dynamic prediction system of river temperature to waste cooling water discharge at 1 km downstream from a thermal power station from real-time to time ahead (2 h) sequentially at the new arrival of each item of river, hydrological, meteorological, power station operational data. It is demonstrated that the DNFLMS modes 1 and 2 shows a better prediction performance and less computation time required, compared to a well-known adaptive neural-fuzzy inference system (ANFIS) and a multi-layer perceptron (MLP) trained with the back propagation (BP) learning algorithm, due to local generalization approach and one-pass learning algorithm implemented in the DNFLMS. It is shown that the DNFLMS mode 1 is that it can be used for an online modeling task without a large amount of training set required by the off-line learning algorithm of MLP-BP and ANFIS. The integration of the DNFLMS mode 2 with a nonlinear feature extraction technique shows that it can improve model generalization capability and reduce model development time by eliminating iterative procedures of model construction using a stopping criterion in training and the quantity of required available data in training given by the GT.

Chilean wine varietal classification using quadratic Fisher transformation

This paper deals with the Chilean red wine varietal classification problem. The problem is solved here by using one of the simplest statistical classification methods based on quadratic discriminant analysis (QDA) together with a new recently introduced nonlinear feature extraction technique called quadratic Fisher transformation. Classification is based on liquid chromatograms of polyphenolic compounds present in wine samples, obtained from a high performance liquid chromatograph with diode alignment detector. For comparison purposes three other feature extraction methods are studied: linear Fisher transformation, Fourier transform and wavelet transform, maintaining QDA as classification scheme. From experimental results it is possible to conclude that when using quadratic discriminant analysis as classification method, the percentage of correct classification was improved from 91% (obtained for the case of wavelet extraction) to 99% when employing quadratic Fisher transformation as feature extraction method.