1- Introduction

2- Experimental setup

3- Cross wavelet transform

4- Wavelet and scales selection

5- Methodology

6- Results

7- Conclusions

References

Abstract

This paper presents a new wavelet analysis approach in partial discharges cable joint measurements in noisy environments. The proposed technique uses the Cross Wavelet Transform (XWT) to separate PD signals from noise and external disturbances in partial discharges measurements in cable joints using two opposite polarity sensors. The partial discharge measurements were performed during impulse and superimposed voltages, leading to a huge amount of noise and pulse shaped external disturbances. The XWT foundations, the experimental setup and the XWT methodology proposed are presented together with the results of the recognition of PD originated in the cable joint. In the experiments, 51,898 signals were acquired, in which 733 were PD signals from the joint and 51,165 corresponded to noise or external disturbances. The XWT performance was studied, finding that 97% of the PD signals were correctly separated by the technique proposed. The results demonstrate the effectivity of the XWT in separating PD signals from noise and external disturbances in this particular measuring system configuration.

Introduction

Nowadays, Partial Discharges (PD) detection is an essential tool for the diagnosis of high-voltage equipment because of their accuracy to detect and quantify defects and damages in the dielectric insulation, where the detection implies the measurement, acquisition, storage and processing of the PD phenomenon [1]. In general, the most widespread PD detection system is based on electrical measurements, in which the PD signals are acquired in the form of individual or series of electrical pulses. In offline PD cable field tests and in laboratory tests, capacitive coupled sensors installed at the cable ends are used [2,3]. In cable systems, statistically most of the partial discharges come from either the cable terminations or the cable joints, being necessary to locate them by time domain reflectometry techniques. In spite of the PD measurement has been exhaustively researched over the years, the separation of PD pulses from noise is one of the main challenges, especially in online applications. Therefore, noise contamination is one of the significant problems of PD detection [4], because noise, disturbances and interferences can give rise to complex Phase Resolved Partial Discharges (PRPD) patterns or clusters, leading to misleading interpretations [5]. For this reason, several studies [1,4,6–13] have focused on the PD pulses separation and denoising techniques for PD measurements. Among these studies, the wavelet transform has been broadly used because is capable of locating time and frequency components allowing the analysis of aperiodic signals with irregular and transition features, such as the partial discharges [1]. In the wavelet analysis techniques, the Discrete Wavelet Transform (DWT) has been used extensively for denoising PD signals. In general, in the DWT denoising, the wavelet coefficients are calculated for a given signal and then the coefficients are passed through a threshold (soft or hard) and followed by the reconstruction of the signal by taking the inverse wavelet transform of the modified DWT coefficients. However, a major problem that most of these denoising techniques face is the ingress of external interferences having time-frequency characteristics similar to the partial discharge signals (pulse shaped disturbances); for instance, periodic pulse shaped interferences from power electronics or another periodic switching [7], PD and corona discharges from the external power system, electrical pulses from switching operations, lightings, etc. This external noise can cause a false indication of PD activity, jeopardising the PD measurements as a diagnostic tool. To reduce the false indications, in PD measurement systems more than one electrical sensor (HFCTs, UHF antennas, coupling capacitors, etc.) is used, meaning that multiple waveforms are simultaneously acquired, because recording each signal through different sensors may provide extra useful information about the real nature of the waveform recorded. Tools like the correlation and trend analysis can provide the significance of relationships between the signals recorded [14].

1- Introduction

2- Literature review on co-movement of inflation

3- Data

4- Methodology

5- Results and discussion

6- Conclusion

References

Abstract

We investigate the pairwise causality of inflation rates across time and frequencies, inflation cycle synchronization and network structure of causality between five ASEAN countries: Indonesia, Malaysia, The Philippines, Singapore, and Thailand. We draw our empirical results and conclusions by implementing dynamic conditional correlations (DCCs), a wavelet measure of cohesion for inflation cycle evolution assessment, and the spillover network index model of Diebold and Yilmaz [1,2]. We find evidence of time-dependent variation in the strength of co-movement between inflation cycles across countries. Positive network causality between inflation cycles and inflation integration across the ASEAN countries are identified. The lead–lag properties of economic indicators are observed to depend on the cycle’s periodicity. The inflation synchronization is particularly pronounced in Thailand.

Introduction

Global macroeconomic developments have mainly focused on business cycle formation rather than inflation cycle formation, probably due to inflation cycle formation being easily linked with economic activity and growth [3–8]. However, given the intrinsic relationship of dependence and causality between interest rates, inflation cycles, and business cycles [9–15], the focus on macroeconomic developments has gradually shifted towards the analysis of cyclicality in macroeconomic fundamentals. Inflation cycles, in particular, due to the role inflation plays in business cycle formation and development and in causing its fluctuations, with decreasing and increasing business cycle trends varying according to investor response to market innovations, are largely perceived as opportunities for economic growth and as sources of economic risk, simultaneously. It is this complexity of inflation cycles, which underlies interwoven dynamic relationships between global, macro, micro and financial variables in the short- mid- and long runs, and a higher integration of international and regional economies via trade linkages and trade competition, what continues to draw the attention of policy makers, the private sector and the academia. Accordingly, relevant stakeholders are beginning to understand the economic and trade diversification opportunities as well as the risks behind high-long and low-short inflation cycle interdependence, in relation to regional or international trading partners. In addition, interest rate determination is now seen from the perspective of inflation cycles that could be positively or negatively correlated with inflation cycles of substantial trading partners, and it is in this context where the subject of inflation cycle synchronization becomes relevant for macroeconomic decision-making and for determining regional trading agreements [16].

1- Introduction

2- Database

3- Methods

4- Results and discussion

5- Conclusion

References

Abstract

Background and objective: It is challenging to conduct real-time identification of myocardial infarction (MI) due to artifact corruption and high dimensionality of multi-lead electrocardiogram (ECG). In the present study, we proposed an automated single-beat MI detection and localization system using dual-Q tunable Q-factor wavelet transformation (Dual-Q TQWT) denoising algorithm. Methods: After denoising and segmentation of ECG, a fourth-order wavelet tensor (leads × subbands × samples × beats) was constructed based on the discrete wavelet packet transform (DWPT), to represent the features considering the information of inter-beat, intra-beat, inter-frequency, and inter-lead. To reduce the tensor dimension and preserve the intrinsic information, the multilinear principal component analysis (MPCA) was employed. Afterward, 84 discriminate features were fed into a classifier of bootstrapaggregated decision trees (Treebagger). A total of 78 healthy and 328 MI (6 types) records including 57557 beats were chosen from PTB diagnostic ECG database for evaluation. Results: The validation results demonstrated that our proposed MI detection and localization system embedded with Dual-Q TQWT and wavelet packet tensor decomposition outperformed commonly used discrete wavelet transform (DWT), empirical mode decomposition (EMD) denoising methods and vectorbased PCA method. With the Treebagger classifier, we obtained an accuracy of 99.98% in beat level and an accuracy of 97.46% in record level training/testing for MI detection. We also achieved an accuracy of 99.87% in beat level and an accuracy of 90.39% in record level for MI localization. Conclusion: Altogether, the automated system brings potential improvement in automated detection and localization of MI in clinical practice.

Introduction

Myocardial infarction (MI) is defined as myocardial cell death due to prolonged ischemia [1]. As one of the main causes of death and disability, MI is an intractable disease and can result in artery disease. In clinical practice, many techniques, including electrocardiographic (ECG), biochemical markers, imaging and so on, are used to assist in the diagnosis of MI. Among these techniques, the non-invasive ECG, an economic tool, is widely used in MI detection [2,3]. The ECG abnormalities of MI can be observed in the PR segment, the QRS complex, the ST segment or the T wave [1]. However, the diagnosis of MI usually requires multiple ECGs because the ECG signals are time-varying in nature with small amplitude. Manual inspection in clinical practice is not only timeconsuming and strenuous but also leads to inter- and intraevaluator variability [4,5]. Therefore, a computer-aided diagnosis system (CADS) of MI should be developed to realize time-saving and reliable analysis [6–11]. Good quality ECG is a guarantee of reliable CADS, while the ECG signals are often corrupted by noise [12]. The ECG signals are usually mixed with different kinds of artifacts, such as power line interference, muscle artifacts, and baseline drifts. Therefore, it is necessary to remove artifacts by implanting denoising method in CADS.

1- Introduction

2- Methods and materials

3- Results

4- Discussion

5- Conclusion

References

Abstract

Atrial fibrillation (AF) is a common cardiac arrhythmia in clinic. The traditional AF detection using visual inspection by trained physicians is an inefficient and burdensome task. In this paper, we introduce a novel method for the automated AF detection using two-lead electrocardiogram (ECG) signals. We use the wavelet packet transform (WPT) and the correlation function of random process theory to devise an efficient feature extraction strategy for physiological signal analysis, and construct the corresponding histogram. Then, multivariate statistical features based on the correlation among wavelet coefficient series are extracted from the corresponding histogram as the feature set, which is the input to artificial neural network (ANN) classifier for the detection. Moreover, various statistical analyses are performed and some parameter tuning strategies are formulated by fitting the receiver operating characteristic (ROC) curve to ensure the reliability and robustness of the work. To evaluate the classification performance of the algorithm, 10-fold cross-validation is implemented on the MIT-BIH AF database. Compared with some state-of-the-art algorithms, the numerical results prove that our proposed strategy yields superior classification performance. To the best of our knowledge, this is also the first application of random process theory for AF detection, providing great potential in medical diagnosis.

Introduction

Atrial fibrillation (AF), a common cardiac arrhythmia, endangers human health and causes a social burden, which is closely related to the occurrence of stroke, thrombus, even death. More than 6 million people have been affected in the United States and about 90 million people have AF around the world [1]. This trend is expected to increase significantly with age [2]. Therefore, it is essential for public health to combat the harm and prevalence of the disease. Conventionally, AF detection is diagnosed by visual inspection of electrocardiogram (ECG) signals by trained physicians, whichmakes artificialdetectioninefficient andsubjective [3]. A large amount of ECG data have inevitably hindered the efficiency of AF detection. Hence,there is an urgent demand for an automated AF detection mechanism to analyze massive amounts of ECG data, facilitate diagnosis and lighten the burden on physicians [4]. The ECG is one of the most common and powerful tools for diagnosing atrial activity in clinicaltreatment, which mainly reveals the electrical action in the heart of human body [3,5]. The irregularity of RR intervals and absence of P-waves (replaced by rapid, irregular and disordered fibrillatory waves, called f-waves) are two main features of ECG data in AF [6]. Various algorithms based on these two features have been performed to automatically detect AF from ECG data. According to the absence of P-waves, Maji et al. [7] utilized empirical mode decomposition to analyze the denoised ECG signals for the corresponding intrinsic mode functions, and then some statistical parameters of P-waves were derived from these functions to perform classification. Padmavathi and Ramakrishna [8] measured some related coefficients based on the auto-regressive model of ECG signals and the support vector machine (SVM) as well as K-nearest neighbor (KNN) classifier was used to detect AF episodes. Pourbabaee et al.[9] developed a feature learning method using deep convolutional neural networks for the analysis of ECG data. At the same time, more schemes using the irregularity of RR intervals have also been initiated to classify AF signals. Dash et al.

1- Introduction

2- Related work

3- Materials and method

4- Experimental results

5- Discussion

6- Conclusion and future direction

References

Abstract

Epilepsy, a common neurological disorder, is generally detected by electroencephalogram (EEG) signals. Visual inspection and interpretation of EEGs is a slow, time consuming process that is vulnerable to error and subjective variability. Consequently, several efforts to develop automatic epileptic seizure detection and classification methods have been made. The present study proposes a novel computer aided diagnostic technique (CAD) based on the discrete wavelet transform (DWT) and arithmetic coding to differentiate epileptic seizure signals from normal (seizure-free) signals. The proposed CAD technique comprises three steps. The first step decomposes EEG signals into approximations and detail coefficients using DWT while discarding non-significant coefficients in view of threshold criteria; thus, limiting the number of significant wavelet coefficients. The second step converts significant wavelet coefficients to bit streams using arithmetic coding to compute the compression ratio. In the final step, the compression feature set is standardized, whereupon machine-learning classifiers detect seizure activity from seizure-free signals. We employed the widely used benchmark database from Bonn University to compare and validate the technique with results from prior approaches. The proposed method achieved a perfect classification performance (100% accuracy) for the detection of epileptic seizure activity from EEG data, using both linear and non-liner machine-learning classifiers. This CAD technique can thus be considered robust with an extraordinary detection capability that discriminates epileptic seizure activity from seizure-free and normal EEG activity with simple linear classifiers. The method has the potential for efficient application as an adjunct for the clinical diagnosis of epilepsy.

Introduction

Epilepsy is a common brain disorder that affects people of all ages. It is a chronic neurological disorder in which recurrent seizures occurs due to abnormal neuronal activities within the human brain and affects the sensorium, mood and/or movement of the human body [1]. World Health Organization statistics indicate that approximately 50 million people currently live with epilepsy worldwide and an estimated 2.4 million people are diagnosed with epilepsy each year [2]. The incidence of the malady is higher in developing countries;i.e., between 7 and 14 per 1000 people. Treatment mostly comprises antiepileptic drugs and/or surgery [3]. The electroencephalogram (EEG) commonly detects seizure activity as it reflects electrophysiological conditions of the brain at a given time [4] and is widely used for diagnostic due to its low cost. EEG signals, enhanced with physiological and pathological data, are employed to evaluate and assess the treatment and progress of epileptic patients. Typically, clinicians evaluate EEG signals for three types of activity: (i) normal EEG activity that records healthy subjects with eyes open or closed; (ii) inter-ictal/seizure-free EEG activity that may contain small spikes and/or subclinical seizures that occur between two clinical episodes in epileptic patients; (iii) and ictal EEG activity containing sudden spikes.

In the last two decades, both Empirical Mode Decomposition (EMD) and Intrinsic Time-Scale Decomposition (ITD) algorithms deserved a variety of applications in various fields of science and engineering due to their obvious advantages compared to conventional (e.g. correlation- or spectral-based analysis) approaches like the ability of their direct application to non-stationary signal analysis. However, high computational complexity remains a common drawback of these otherwise universal and powerful algorithms. Here we compare similarly designed signal analysis algorithms utilizing either EMD or ITD as their core functions. Based on extensive computer simulations, we show explicitly that the replacement of EMD by ITD in several otherwise similar signal analysis scenarios leads to the increased noise robustness with simultaneous considerable reduction of the processing time. We also demonstrate that the proposed algorithms modifications could be successfully utilized in a series of emerging applications for processing of non-stationary signals.

Introduction

Majority of the observational signals considered in various fields of knowledge are non-stationary indicated by the variability of their statistical characteristics over time. In most practical scenarios, the stationarity assumption is validated for the first- and second-order moments only indicated by the time independence of both average and variance as well as the autocorrelation function having only a single time scale argument. Existing approaches to the non-stationary signal analysis and processing have several significant drawbacks. For example, the widely used classical Fourier analysis, due to its relative calculation simplicity and fast computational algorithms availability, immediately began to overwhelm all other signal analysis methods. Despite the fact that the Fourier transform is performed under very general assumptions such as the Dirichlet boundary conditions and absolute integrability, there are several significant limitations on the signals for which it is calculated [1]. Fourier analysis has been originally suggested for strictly periodic functions that could be directly expanded in a Fourier series, i.e. represented by a superposition of multiple harmonic functions. Otherwise, the frequency domain based analysis may lead to incorrect results interpretation. It is also necessary that signals exhibiting the stationarity property for their certain characteristics such as average values and instant frequencies [1].

1. Introduction

2. Enhancement Algorithm

3. Experimental Results

4. Conclusion

References

In order to improve the contrast and sharpness of traffic sign images obtained under low light natural environment, we propose an improved enhancement method based on discrete wavelet transform to improve image contrast. We convert the original RGB image to the HSV color space, and use the discrete wavelet transform (DWT) to decompose the luminance component (V). In the low-frequency component use multi-scale Retinex algorithm estimate the illuminance to enhance the contrast of images, the high-frequency component enhances the detail information through the multi-scale detail boosting method. Finally, adjust the saturation component (S) by a piecewise exponential transformation method to make the image color more suitable for human observation. Experimental results demonstrate that our method can better display image details while reducing the halo effect, and effectively improve the contrast and sharpness of low-light images compared with existing algorithms through subjective and objective analysis.

Introduction

As an important part of the Advanced Driver Assistance System (ADAS), Traffic Sign Recognition (TSR) can recognize the traffic sign information in real time and provide it to the driver, thus reducing driver’s driving pressure, effectively ensuring driving safety, and avoiding traffic accidents [1]. However, the traffic sign images collected by the in-vehicle devices under low illumination conditions generally have problems such as a decrease in global contrast, content blurring, or loss of details [2]. This phenomenon has an adverse effect on the subsequent detection and recognition of traffic signs, Therefore, how to enhance the visual effect of traffic sign images under low light environment and highlight useful information of images has become an urgent problem to be solved.

I. INTRODUCTION

II. DAMAGE DETECTION METHODOLOGY

III. DISPERSION CURVES

IV. FE SIMULATION MODEL

V. DAMAGE DETECTION STRATEGY

VI. DAMAGE LOCATION ESTIMATION

VII. DAMAGE SIZE ESTIMATION

VIII. INFLUENCE OF NOISE ON DAMAGE ESTIMATION

IX. DISCUSSION OF RESULTS

X. CONCLUSION

REFERENCES

The damage detection method for composite laminates introduced in this research uses piezoelectric (lead zirconate titanate, PZT) transducers to excite/sense the Lamb wave signals. The complicated wave signals scattered by damage are accurately processed using a continuous wavelet transformation (CWT) based on the Gabor wavelet. The transducers are arranged on a composite laminate in the form of a network of square detection cells and triangular subcells. The damage location is estimated using the concept of centroid in two-stage detection method. The first stage detection is carried out by exciting a transducer at the center of each detection cell to locate the damaged cell and subcell. The damage localization is improved by exciting an additional transducer at the corner of the damaged subcell during the second stage detection. The damage size is then quantitatively estimated using cubic spline curve (CSC) and elliptical parametric (EP) methods based on the damage edge points. The damage location is estimated in two detection stages for high-accuracy because the damage edge points are calculated with reference to the estimated location of the damage. The arrangement of transducers and signal processing technique remain the same at all the stages of damage detection. Results from previous detection stages contribute to the improvement of damage detection in the subsequent stages. The size of detection cell plays a crucial role in designing the detection stages, and the proposed method can accurately quantify both location and size of the damage in composite laminate.

INTRODUCTION

The properties of composite materials can be tailored to exhibit superior properties such as high strength to weight ratio, high modulus and high resistance to aggressive environmental conditions. Composite materials are increasingly employed to achieve the desired performance in wide range of applications from various industries like aerospace, automotive, naval and wind turbine etc. However, the mechanical properties and structural integrity of composite structures can be compromised by various types of fabrication and in-service defects. A small damage in composite structure can grow and lead to the catastrophic failure of the entire structure. Therefore, a continuous monitoring of the structural components, for damage detection at an early stage, is required for efficient repair and maintenance. Researchers have developed many computational and intelligent techniques for structural health monitoring (SHM) of composites [1]. The SHM techniques based on Lamb waves are commonly used for damage detection in composite laminates. The Lamb waves are highly sensitive to small imperfections, and have the ability to propagate relatively a long distance with low attenuation. The change in the Lamb wave propagation due to forward and backward scattering from damage can be analyzed to evaluate the damage. The PZT (lead zirconate titanate) transducers are frequently used to excite and sense the Lamb wave signals because they are small, lightweight, consume low power, and produce a frequency response in wide region [2].

I. INTRODUCTION

II. PRELIMINARIES

III. WAVELET PACKET BASED TEXTURE RETRIEVAL

IV. WAVELET PACKET BASIS SELECTION

V. SUMMARY OF THE MODELING STAGE

VI. EXPERIMENTAL ANALYSI

VII. DISCUSSION: CONNECTION WITH CNN

VIII. CONCLUSION AND FUTURE WORK

APPENDIX WAVELET PACKETS ANALYSIS

ACKNOWLEDGMENT

REFERENCES

Wavelet Packets (WPs) bases are explored seeking new discriminative features for texture indexing. The task of WP feature design is formulated as a learning decision problem by selecting the filter-bank structure of a basis (within a WPs family) that offers an optimal balance between estimation and approximation errors. To address this problem, a computationally efficient algorithm is adopted that uses the tree-structure of the WPs collection and the Kullback-Leibler divergence as a discrimination criterion. The adaptive nature of the proposed solution is demonstrated in synthetic and real data scenarios. With synthetic data, we demonstrate that the proposed features can identify discriminative bands, which is not possible with standard wavelet decomposition. With data with real textures, we show performance improvements with respect to the conventional Wavelet-based decomposition used under the same conditions and model assumptions.

INTRODUCTION

In the current information age, we have access to unprecedented sources of digital image content. Consequently, being able to index and organize these documents based solely on the content extracted from the signals without relying on metadata or expensive human annotations has become a central problem [1]–[21]. In this context, an important task in image processing is texture retrieval. This problem has been richly studied over the last two decades with different frameworks and approaches [3]–[21], including, more recently, deep learning approaches [22], [22]–[26], [26]–[29]. In a nutshell, the texture retrieval problem can be formulated in two stages. The first stage, feature extraction (FE), implies the creation of low-dimensional descriptions of the image (i.e., the dimensionality reduction phase) with the objective of capturing the semantic high-level information that discriminates relevant texture classes. The second stage proposes a similarity measure (SM) on the feature space to compare and organize the images in terms of their signal content. For the FE stage, the Wavelet transform (WT) has been widely adopted as a tool to decompose and organize the signal content in sub-spaces associated with different levels of resolution (or scale) information [30], [31]. Based on this sub-space decomposition, energy features have been used as a signature that represents the salient texture attributes for texture indexing [32].

Eye and systemic diseases are known to manifest themselves in retinal vasculature. Segmentation of retinal vessel is one of the important steps in retinal image analysis. A simple unsupervised method based on Gabor wavelet and Multiscale Line Detector is proposed for retinal vessel segmentation. Vessels are enhanced by linear superposition of first scale Gabor wavelet image and complemented Green channel. Multiscale Line Detector is used to segment the blood vessels. Finally, a simple post processing scheme based on median filtering is deployed to remove false positives. The proposed scheme was evaluated with publicly available datasets called DRIVE, STARE and HRF, obtaining an accuracy of 0.9470, 0.9472, and 0.9559, and a sensitivity of 0.7421, 0.8004, and 0.7207, respectively. These results are comparable to the state-of-the-art methods, albeit with a simpler approach.

Introduction

One of the important tasks in diagnosing different medical conditions such as diabetic retinopathy, cardiovascular diseases, and stroke is the segmentation of blood vessels in color medical images. To this end, different strategies have been devised. The strategies can be roughly grouped into i) multiscale, ii) matched filtering, iii) mathematical morphology, iv) hierarchical, v) model and vi) deep learning approach [1]. Furthermore, they can also be categorized into supervised and unsupervised algorithm. The prominent strategies based on multiscale are [2], [3]. Soares et al. [2] used Gabor wavelet transform with four scales (2,3,4,5) to account for different width sizes of blood vessel, and supervised classification. Nyugen et al. [4] proposed blood vessel segmentation using a multi-scale line detection based technique. The approach is an extension of the scheme based on single scale line detector and support vector machine [5]. Examples of filter based approach are [6]–[8] and mathematical morphology based are [9]–[11]. Retinal vessel segmentations based on hierarchical detections are [12], [13], model based approaches [14]–[16], and deep learning based schemes [13], [17]–[21]. One major challenge in retinal image analysis especially for accurate vessel detection is low and varying contrast. A method based on Gabor wavelet and multi-scale line detector is being proposed here. The Gabor wavelet transform presents high frequency precision in low frequencies and high spatial precision in high frequencies. In other words, the transform is suitable for detecting edges and other singularities in the image [22], [23].

Social networks generate enormous amounts of visual data. Mining of such data in recommender systems is extremely important. User profiling is carried out in recommender systems to build the holistic persona of the user. Identification and grouping of images in these systems is carried out using face recognition. It is one of the most appropriate biometric features in such situations. Ever since the first use of face recognition in security and surveillance systems, researchers have developed many methods with improved accuracy. Face recognition under variant illumination is still an open issue and diverging facial expressions reduces the accuracy even further. State of the art methods produced an average accuracy of 90%.In this study, a computationally intelligent and efficient method based on particle swarm optimization (PSO) is developed. It utilizes the features extracted from texture and wavelet domain. Discrete Wavelet Transform provides the advantage of extracting relevant features and thereby reducing computational time and an increase in recognition accuracy rate. We apply particle swarm optimization technique to select informative wavelet sub-band. Furthermore, the proposed technique uses Discrete Fourier Transform to compensate the translational variance problem of the discrete wavelet transform. The proposed method has been tested on the CK, MMI and JAFFE databases. Experimental results are compared with existing techniques and the results indicate that the proposed technique is more robust to illumination and variation in expressions, average accuracy obtained over the CK, MMI and JAFFE datasets is 98.6%, 95.5%, and 98.8% respectively.

Introduction

Face Recognition (FR) is a biometric solution that is used for the identification or authentication of a human from a video or image source. It has been successfully utilized in a variety of domains. Key application areas of facial recognition include augmented reality, retailed marketing industry, gaming, security, forensics, video conferencing, smart meetings, visual surveillance and anti-terrorism. It is a process in which the unique facial characteristics of a person are matched with the templates stored in a facial database [1]. Finding an automated solution for the face recognition problem is not a trivial task due to various factors including variable lighting effects [2], different facial expressions and postures [3] in different images of the same person. Face recognition techniques can be classified into three groups; feature-based, holistic, and hybrid of these two [1]. In face recognition process, facial features’ extraction process plays an important role and it involves a number of decisions like the selection of appropriate features, description, and Special Issue on Computational Intelligence Paradigms in Recommender Systems and Online Social Networks representation of these features. The fundamental goal is to represent patterns with the most significant features from the feature set yet with minimal loss of important facial information [4]. The FR methodology is highly dependent on the quality of such feature selection that serves as an input to the classifier. A feature extractor of good quality must be capable of extracting the discriminant features in unconstrained situations like outdoors, variable illumination, variation in poses, and facial expressions etc. Majority of the holistic feature extraction technique uses a combination or variations of Neural Networks, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) [5] etc. PCA-based face recognition has the advantages of simplicity and efficiency but is vulnerable in the presence of different lighting conditions, translation-variance, and different poses [6]. LDA-based techniques are suitable candidates for feature extraction and dimensionality reduction. These are generally considered better than the conventional PCA-based techniques and are also able to solve the illumination problem [7]. Transform domain-based techniques extract the major features of a facial image by performing the transformation from one coordinate system to a new orthogonal coordinate system resulting in the reduction of dimensionality and compact representation of data.

Broadband environmental excitations from wind, ocean wave and earthquakes are especially dangerous for flexible tall structures such as wind turbines, towers or cable-stayed bridges. Liquid Dampers have been studied for almost thirty years in terms of their capability for suppression of vibration in such structures. The work presented in this paper focuses on the Tuned Liquid Column Damper, both open and sealed, and the identification of its time-varying properties of nonlinear damping, frequency and air pressure identification. Experimental tests are conducted on a full scale model of the damper which is subjected to both white noise and harmonic excitation by means of a hydraulic shaker. Exponential decay of the displacement of the liquid column was measured and analysed. The identification procedure was conducted step-wise, first, mode separation with the use of Continuous Wavelet Transform was carried out and then identification of the instantaneous damping ratio for the first mode of vibration was performed. Results indicate that the damping ratio is nonlinear, timevarying and depends on the level of vibration. The air pressure data in the sealed TLCD was also recorded and analysed to reveal the nonlinear nature of the pressure change and the presence of higher odd harmonics.

Introduction

Liquid dampers have been successfully utilized in practise to supress amplitudes of structural vibration under wind, wave and earthquake excitation. The concept of using fluid to stabilize the rolling of ships was developed in the 1860s [1]. This damping technology is known by the name of anti-roll-tanks and widely used in naval architecture [2]. As compared to sloshing [3] or rectangular dampers [4], Tuned Liquid Column Dampers (TLCDs), which dissipate energy by the movement of an oscillatory column of liquid through orifice(s) provided in the cross section of a U-shaped container, have attracted special attention due to their high volumetric efficiency with respect to given amount of liquid, consistent behaviour across a wide range of excitation levels and a damping mechanism that can be quantified in a definite manner. The theoretical concept of the TLCD was developed in [5] and experimentally verified in [6]. Also its nonlinear mathematical description was provided and the relationship between head-loss coefficient and liquid damping validated. Since then, considerable research work has been carried out on the characterization and performance of TLCDs [7–14] including investigations on applicability to control of short period structures, wind turbines and on the impact of soil-structure interaction. Since a TLCD is described by a relatively simple mathematical model, it is amenable for semi-active and active control. Some studies on TLCD, acting as an active vibration damper, were carried out in [15,16]. Experimental study for calculating the effectiveness of semiactive LCD for wind-induced vibration was carried out in [17]. Effectiveness of the semi-active LCD system and the hybrid viscous damper LCD control system for the suppression of wind-induced motion of highrise buildings was examined in [18]. Attempts to achieve enhanced TLCD performance by using Electro-Rheological fluid (ER fluid) or Magneto-Rheological fluid (MR fluid) were presented in [19–21]. Experimental and theoretical investigations on the equivalent viscous damping of structures with TLCD having MR-fluids was carried out in [14,22]. Some researchers considered the vertical limbs of the TLCD to be sealed and utilized the air-spring effects in the sealed U-tube to extend the applicability of the control device to the high frequency range [23]. Sealed TLCDs were also studied for the control of wind induced multi-modal lateral and torsional vibration of long span cable stayed bridges by [24] and for the seismic vibration control of steel jacket platforms by [25]. Previous research studies in the field are related mainly to the design and control aspects of TLCDs. Identification of vibration/modal parameters has attracted much less attention. This is particularly relevant to the identification and characterisation of the time-varying behaviour of these parameters. The present paper aims to model and experimentally identify nonlinear, time-varying damping in conventional TLCDs and nonlinear behaviour of air pressure in sealed TLCDs. The structure of the paper is as follows. Two simple mathematical models associated with the analysed nonlinear behaviour of the TLCD are presented in Section 2. The identification procedure, based on wavelet analysis, is briefly described in Section 3. Experimental work and results are presented in Section 4. Finally, the paper is concluded in Section 5.

Abstract

The spectral-based features, typically used in Automatic Speech Recognition (ASR) systems, reject the phase information of speech signals. Thus, employing extra features, in which the phase of the signal is not rejected, may fill this gap. Embedding the speech signal in the Reconstructed Phase Space (RPS) and then extracting some useful features from it, is a recently considered approach in this field. In this paper, we will follow this approach by evaluating some useful features from the Recurrence Plot (RP) of the embedded speech signals in the RPS; the proposed features are evaluated via applying a two-dimensional wavelet transform to the resulted RP diagrams. The proposed features are examined in an ASR task alone and in combination with the traditional Mel-Frequency Cepstral Coefficients (MFCC). For the second case, using English TIMIT corpus, 3.94% absolute classification accuracy improvement in the phoneme recognition accuracy rate, against using only the MFCC features is gained.

Introduction

In recent decades, a variety of linear models for speech coding, synthesis, and recognition with acceptable performances have been introduced. In this way, many types of research achieved improvement in the field of speech recognition by employing novel methods [1,2] or the detection of mispronunciation using Hidden Markov model [3]; however, there are nonlinear aerodynamic phenomena in the human speech production system which generally could not be included in linear models [4]. Therefore, nonlinear methods could potentially provide effective computational models to extract acoustic features which are useful for the nonlinear phenomena detection [4]. Furthermore, some recent studies have shown that utilizing nonlinear characteristics may improve the performances of the ASR systems [5].

Usual ASR systems exploit frequency domain features like Mel frequency cepstral coefficients [6]. The traditional frequency domain methods typically extract only the first and second order properties from the spectral patterns of speech signals [7]. However, there are many signals produced via nonlinear differential equations that have wide spectral characteristics [8]. In such cases, the frequency domain techniques are deficient, because it is impossible to dissociate the information of such a signal only in the frequency domain [5].