Conventional Feature Extraction Methods Research Articles

Band Clustering-Based Feature Extraction for Classification of Hyperspectral Images Using Limited Training Samples

Feature extraction plays a central role in classification of hyperspectral data. We propose a clustering-based feature extraction (CBFE) method in this letter. The proposed method is supervised and only needs to calculate the first-order statistics. Thus, CBFE has better performance than some popular supervised feature extraction methods such as linear discriminant analysis, generalized discriminant analysis, and nonparametric weighted feature extraction in small sample size situation. In addition, CBFE works better than unsupervised approaches such as principal component analysis in classification applications. CBFE considers a vector associated with each band that is composed by the mean values of all classes in that band. Then, a clustering method such as k-means is run to group the similar bands in one cluster. The selected number of clusters is equal to the number of extracted features. Experiments carried out on two different hyperspectral data sets demonstrate that the CBFE has better performance in comparison with some conventional feature extraction methods.

(Semi-) Supervised Probabilistic Principal Component Analysis for Hyperspectral Remote Sensing Image Classification

In this paper, we have applied supervised probabilistic principal component analysis (SPPCA) and semi-supervised probabilistic principal component analysis (S <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> PPCA) for feature extraction in hyperspectral remote sensing imagery. The two models are all based on probabilistic principal component analysis (PPCA) using EM learning algorithm. SPPCA only relies on the labeled samples into the projection phase, while S <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> PPCA is able to incorporate both the labeled and unlabeled information. Experimental results on three real hyperspectral images demonstrate the SPPCA and S <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> PPCA outperform some conventional feature extraction methods for classifying hyperspectral remote sensing image with low computational complexity.

Fuzzy-based discriminative feature representation for children's speech recognition

Automatic recognition of the speech of children is a challenging topic in computer-based speech recognition systems. Conventional feature extraction method namely Mel-frequency cepstral coefficient (MFCC) is not efficient for children's speech recognition. This paper proposes a novel fuzzy-based discriminative feature representation to address the recognition of Malay vowels uttered by children. Considering the age-dependent variational acoustical speech parameters, performance of the automatic speech recognition (ASR) systems degrades in recognition of children's speech. To solve this problem, this study addresses representation of relevant and discriminative features for children's speech recognition. The addressed methods include extraction of MFCC with narrower filter bank followed by a fuzzy-based feature selection method. The proposed feature selection provides relevant, discriminative, and complementary features. For this purpose, conflicting objective functions for measuring the goodness of the features have to be fulfilled. To this end, fuzzy formulation of the problem and fuzzy aggregation of the objectives are used to address uncertainties involved with the problem.The proposed method can diminish the dimensionality without compromising the speech recognition rate. To assess the capability of the proposed method, the study analyzed six Malay vowels from the recording of 360 children, ages 7 to 12. Upon extracting the features, two well-known classification methods, namely, MLP and HMM, were employed for the speech recognition task. Optimal parameter adjustment was performed for each classifier to adapt them for the experiments. The experiments were conducted based on a speaker-independent manner. The proposed method performed better than the conventional MFCC and a number of conventional feature selection methods in the children speech recognition task. The fuzzy-based feature selection allowed the flexible selection of the MFCCs with the best discriminative ability to enhance the difference between the vowel classes.

Convolutive Bottleneck Network with Dropout for Dysarthric Speech Recognition

In this paper, we investigate the recognition of speech produced by a person with an articulation disorder resulting from athetoid cerebral palsy. The articulation of the first spoken words tends to become unstable due to strain on speech muscles, and that causes degradation of speech recognition. Therefore, we propose a robust feature extraction method using a convolutive bottleneck network (CBN) instead of the well-known MFCC. The CBN stacks multiple various types of layers, such as a convolution layer, a subsampling layer, and a bottleneck layer, forming a deep network. Applying the CBN to feature extraction for dysarthric speech, we expect that the CBN will reduce the influence of the unstable speaking style caused by the athetoid symptoms. Furthermore, we also adopt dropout in the output layer since automatically assigned labels to the dysarthric speech are usually unreliable due to ambiguous phonemes uttered by the person with speech disorders. We confirmed its effectiveness through word recognition experiments, where the CNN based feature extraction method outperformed the conventional feature extraction method.

Analysis of Human Electrocardiogram for Biometric Recognition Using Analytic and AR Modeling Extracted Parameters

In this paper, a new approach for automatic analysis of single lead ECG for human recognition is proposed and evaluated. Following the pre-processing step, the ECG stream is partitioned into separate windows where each window includes single beat of ECG signal. After successful QRS detection, various temporal, amplitude and AR coefficients are extracted and used as an input to a classifier in order to identify the individuals. In this work, proposed system has been tested using records from three different publicly available ECG databases. Signal pre-processing techniques, applied parameter extraction methods and some intermediate and final classification results are presented in this paper. that PAW yields equivalent performance in terms of accuracy compared to conventional temporal and amplitude feature extraction methods. Even though, PAW is complicated process which needs powerful digital signal processors to overcome the time delay. In this paper, a new approach for automatic analysis of single lead electrocardiogram (ECG) for human recognition and individual identification is proposed. This approach depends on on analytic (Amplitude, Time and Width) and modelling (AR) features extracted from the ECG beat. Obtained results indicate high level of accuracy and shorter processing time needed to identify the individuals. Eighteen analytic and modelling features are extracted to identify individuals and k nearest neighbour (knn) classification algorithm applied in order to classify those features and evaluate the proposed approach. ECG feature selection and extraction using AR modelling has recently been used (9) resulting in accurate classification of various arrhythmia and ventricular arrhythmia conditions. The remainder of this paper is organized as follows. Section II gives a brief description of the techniques used in the pre-processing phase to clean ECG signals of noise and other artefacts. Section III provides a review of QRS detection methods used in this work. Feature selection and extraction methods are discussed in Section IV whilst Section V contains experimental results and discussion of those results. Conclusions are presented in Section VI.

Camera calibration by using fringe patterns and 2D phase-difference pulse detection

Camera calibration is the most essential and usually the first step in computer vision applications. The results highly depend on the accuracy of the feature extraction. Conventional feature extraction methods suffer from perspective and lens distortion. In this paper, a novel feature extraction method is proposed by using fringe patterns groups as calibration target. Each group comprises six sinusoidal, in which, three are used for calculating vertical phases and the other three for horizontal phases. A three-step phase shift algorithm is used for wrapped phase calculation. Then, feature points are detected with a 2D phase-difference pulse detection method and refinement is done by simple interpolation. Finally, camera calibration is done by using these features as control points, experimental results indicated the proposed method accurate and robust.

New Feature Extraction Method Based on Contourlet Transform for Banknote Classification

In order to improve the performance of the banknote classification, new banknote image feature extraction method is proposed in this paper. The contourlet transform is applied to the original banknote image which is obtained by image contact sensor.The statistical characteristics of transformed image in the contourlet domain are analyzed. The statistical characteristics which can perfectly reflect the banknote image texture information are used as feature vector for banknote classification. The experimental results show that the proposed method can obtain higher recognition compared with other conventional banknote image feature extraction methods.

Comparative Analysis on Scene Image Classification using Selected Hybrid Features

A comparative analysis on image classification is accomplished on scene image feature set by using various existing classifiers. The classification is performed on conventional feature set, hybrid feature set and selected hybrid feature set for classifying the war tanks from the natural scene images. The features are extracted in three ways; conventional feature extraction methods like gray level co-occurrence matrices features & statistical moment’s features; hybrid feature extraction is the combination of color mean, GLCM properties and canny edge count; the proposed selected hybrid feature set. The extracted features are trained and tested with various classifiers like Artificial Neural Network (ANN) using feed forward back propagation algorithm, Support Vector Machines (SVM) using polynomial kernel with p=1, Bayes Net classifier using genetic search and J4.8 decision tree. The results show that classification efficiency of the selected hybrid feature extraction methods (i.e., the combination of GLCM & edge count) surpasses the conventional feature extraction methods in war scene classification problems.

Spatio-spectral filters for low-density surface electromyographic signal classification

In this paper, we proposed to utilize a novel spatio-spectral filter, common spatio-spectral pattern (CSSP), to improve the classification accuracy in identifying intended motions based on low-density surface electromyography (EMG). Five able-bodied subjects and a transradial amputee participated in an experiment of eight-task wrist and hand motion recognition. Low-density (six channels) surface EMG signals were collected on forearms. Since surface EMG signals are contaminated by large amount of noises from various sources, the performance of the conventional time-domain feature extraction method is limited. The CSSP method is a classification-oriented optimal spatio-spectral filter, which is capable of separating discriminative information from noise and, thus, leads to better classification accuracy. The substantially improved classification accuracy of the CSSP method over the time-domain and other methods is observed in all five able-bodied subjects and verified via the cross-validation. The CSSP method can also achieve better classification accuracy in the amputee, which shows its potential use for functional prosthetic control.

Feature Extraction Based on Speech Attractors in the Reconstructed Phase Space for Automatic Speech Recognition Systems

In this paper, a feature extraction (FE) method is proposed that is comparable to the traditional FE methods used in automatic speech recognition systems. Unlike the conventional spectral-based FE methods, the proposed method evaluates the similarities between an embedded speech signal and a set of predefined speech attractor models in the reconstructed phase space (RPS) domain. In the first step, a set of Gaussian mixture models is trained to represent the speech attractors in the RPS. Next, for a new input speech frame, a posterior-probability-based feature vector is evaluated, which represents the similarity between the embedded frame and the learned speech attractors. We conduct experiments for a speech recognition task utilizing a toolkit based on hidden Markov models, over FARSDAT, a well-known Persian speech corpus. Through the proposed FE method, we gain 3.11% absolute phoneme error rate improvement in comparison to the baseline system, which exploits the mel-frequency cepstral coefficient FE method.

Feature Extraction Based on Speech Attractors in the Reconstructed Phase Space for Automatic Speech Recognition Systems

In this paper, a feature extraction (FE) method is proposed that is comparable to the traditional FE methods used in automatic speech recognition systems. Unlike the conventional spectral-based FE methods, the proposed method evaluates the similarities between an embedded speech signal and a set of predefined speech attractor models in the reconstructed phase space (RPS) domain. In the first step, a set of Gaussian mixture models is trained to represent the speech attractors in the RPS. Next, for a new input speech frame, a posterior-probability-based feature vector is evaluated, which represents the similarity between the embedded frame and the learned speech attractors. We conduct experiments for a speech recognition task utilizing a toolkit based on hidden Markov models, over FARSDAT, a well-known Persian speech corpus. Through the proposed FE method, we gain 3.11% absolute phoneme error rate improvement in comparison to the baseline system, which exploits the mel-frequency cepstral coefficient FE method.

Semisupervised Local Discriminant Analysis for Feature Extraction in Hyperspectral Images

We propose a novel semisupervised local discriminant analysis method for feature extraction in hyperspectral remote sensing imagery, with improved performance in both ill-posed and poor-posed conditions. The proposed method combines unsupervised methods (local linear feature extraction methods and supervised method (linear discriminant analysis) in a novel framework without any free parameters. The underlying idea is to design an optimal projection matrix, which preserves the local neighborhood information inferred from unlabeled samples, while simultaneously maximizing the class discrimination of the data inferred from the labeled samples. Experimental results on four real hyperspectral images demonstrate that the proposed method compares favorably with conventional feature extraction methods.

Engine fault diagnosis based on a morphological neural network using a morphological filter as a preprocessor

Feature extraction and faults classification are the two most significant issues involved in the field of mechanical fault diagnosis problems. In this work, we address these two problems using mathematical morphology and non-negative matrix factorization. In particular, we present a novel engine fault diagnosis scheme utilizing the averaged multi-scale morphological filter to enhance the vibration signals, non-negative matrix factorization to characterize the signals, and a constructive morphological neural network to classify the engine operating states. Eight engine running states including the healthy state and seven defective states are tested in an engine experiment rig to evaluate the presented fault diagnosis scheme. Conventional feature extraction methods as well as classifiers popularly used in the literature are also employed as a comparison. The experimental results indicate the proposed approach to be an effective and efficient scheme for detection of the intelligent faults of engines.

Discrimination and description of repetitive patterns for enhancing the performance of feature-based recognition

Conventional object retrieval or recognition methods based on feature matching sometimes fail when an object contains repetitive patterns, because features from repetitive patterns are too similar to each other. Specifically, when there arise many similar features in a query object due to repetitive patterns, they are usually not matched to the ones at the same positions of the reference object. Hence, the matching pairs between the query and reference image do not appear regular and thus homography estimation fails. In case when we use “nearest neighbor distance ratio” as a matching criterion, where enough distinction between the matched pairs should be secured, matching also fails due to similarity of features. In this paper, we propose a new feature matching strategy to alleviate this problem by discriminating repetitive patterns from the other salient ones and also by developing a way of utilizing the patterns for robust feature matching. Specifically, we first apply a conventional feature extraction method to a given image. Then we cluster features based on their similarity, i.e., we design a classifier that tells whether a feature is from a repetitive pattern or from a salient structure. For the effective use of repetitive patterns, we define a new descriptor based on support vector data description (SVDD) for describing clusters of similar features. In other words, a set of features from a pattern is defined to be a new feature in terms of its center and radius. For object recognition, the homography is found over the salient features by excluding repetitive features at first, which is then validated and refined by the repetitive patterns. The proposed method is tested with examples of matching buildings with repetitive patterns, and it is shown to be robuster and more reliable than the conventional methods.

Recognition of a Driver's Gaze for Vehicle Headlamp Control

In this paper, we propose a novel method for gaze recognition of a driver coping with rotation of a driver's face. Frontal face images and left half profile images were separately trained using the Viola-Jones (V-J) algorithm to produce classifiers that can detect faces. The right half profile can be detected by mirroring the entire image when neither a frontal face nor a left half profile was detected. As an initial step, this method was used to simultaneously detect the driver's face. Then, we applied a regressional version of linear discriminant analysis (LDAr) to the detected facial region to extract important features for classification. Finally, these features were used to classify the driver's gaze in seven directions. In the feature extraction step, LDAr tries to find features that maximize the ratio of interdistances among samples with large differences in the target value to those with small differences in the target value. Therefore, the resultant features are more fitted to regression problems than conventional feature extraction methods. In addition to LDAr, in this paper, a 2-D extension of LDAr is also developed and used as a feature extraction method for gaze recognition. The experimental results show that the proposed method achieves a good gaze recognition rate under various rotation angles of a driver's head, resulting in a reliable headlamp control performance.

Electrocardiogram (ECG) Biometric Authentication Using Pulse Active Ratio (PAR)

Biometric authentication is a one-to-one verification process. A successful biometric system must be capable of avoiding a fraudulent claim while at the same time ensuring the privacy of individuals. This paper presents a novel framework for using the electrocardiogram (ECG) as a biometric system for human authentication. A new feature extraction technique known as pulse active ratio (PAR) is derived and used to generate novel ECG feature vectors. The proposed method is validated by experiments on 112 subjects, performing 9800 ECG comparisons providing a 10% improvement when compared to conventional temporal and amplitude feature extraction methods.

Kernel discriminant analysis for regression problems

In this paper, we propose a nonlinear feature extraction method for regression problems to reduce the dimensionality of the input space. Previously, a feature extraction method LDAr, a regressional version of the linear discriminant analysis, was proposed. In this paper, LDAr is generalized to a nonlinear discriminant analysis by using the so-called kernel trick. The basic idea is to map the input space into a high-dimensional feature space where the variables are nonlinear transformations of input variables. Then we try to maximize the ratio of distances of samples with large differences in the target value and those with small differences in the target value in the feature space. It is well known that the distribution of face images, under a perceivable variation in translation, rotation, and scaling, is highly nonlinear and the face alignment problem is a complex regression problem. We have applied the proposed method to various regression problems including face alignment problems and achieved better performances than those of conventional linear feature extraction methods.

Elongated Radial Basis Function for Nonlinear Representation of Face Data

Recently, subspace analysis has raised its performance to a higher level through the adoption of kernel-based nonlinearity. Especially, the radial basis function, based on its nonparametric nature, has shown promising results in face recognition. However, due to the endemic small sample size problem of face data, the conventional kernel-based feature extraction methods have difficulty in data representation. In this paper, we introduce a novel variant of the RBF kernel to alleviate this problem. By adopting the concept of the nearest feature line classifier, we show both effectiveness and generalizability of the proposed method, particularly regarding the small sample size issue.

A novel gait recognition analysis system based on body sensor networks for patients with Parkinson's disease

Gait analysis of human plays a significant role in maintaining the well-being of our mobility and healthcare, and it can be used for various e-healthcare systems for fast medical prognosis and diagnosis. In this paper, we have developed a novel body sensor network-based recognition system to identify the specific gait pattern of Parkinson's disease (PD). Firstly, a BSN with 16-nodes is used to acquire the gait information from the PD patients. Then, an algorithm is developed based on local linear embedding (LLE) to extract and recognise the gait features. Experiments demonstrate the effectiveness of proposed scheme. The results show that the proposed scheme has a recognition rate of about 95.57% for gait patterns of PD, which is higher than the conventional PCA feature extraction method. The proposed system can identify PD patients from normal people and by their gait map with high reliability and appears a promising aid in the diagnosis of the Parkinson's disease.

Feature extraction based on subspace methods for regression problems

In this paper, we propose a couple of new feature extraction methods for regression problems. The first one is closely related to the conventional principle component analysis (PCA) but unlike PCA, it incorporates target information in the optimization process and try to find a set of linear transforms that maximizes the distances between points with large differences in target values. On the other hand, the second one is a regressional version of linear discriminant analysis (LDA) which is very popular for classification problems. We have applied the proposed methods to several regression problems and compared the performance with the conventional feature extraction methods. The experimental results show that the proposed methods, especially the extension of LDA, perform well in many regression problems.