Traditional Spectral Analysis Methods Research Articles

Unsupervised Feature Selection via Neural Networks and Self-Expression with Adaptive Graph Constraint

Unsupervised feature selection (UFS), which selects the most important feature subset and eliminates the unnecessary information for the upcoming data analysis, is a significant problem in machine learning and has been explored for years. Most UFS methods map features into a pseudo label space by multiplying a projection matrix constrained with sparsity to learn the mapping from the features to the labels. However, the mapping relationship is usually not linear, and linear regression may result in a suboptimal selection. To address this issue, we propose a novel UFS method, called neural networks embedded self-expression (NNSE). NNSE replaces the linear regression of traditional spectral analysis methods with neural networks to learn the pseudo label space. Besides, we embed neural networks into the self-expression model to improve the representative ability by preserving the local structure with an adaptive graph regularization module. Then we propose an efficient alternative iterative algorithm to solve the proposed model. Experimental results on 8 public datasets show NNSE outperforms the other state-of-the-art methods. Moreover, experimental results are also presented to show the convergence of the proposed method. The source code is available at: https://github.com/misteru/NNSE.

Hyperspectral image feature region of solution composition analysis method based on multidimensional spectra

Redundant information is inevitable when acquisition of high-dimension spectra, which contribute to the overfitting.In order to effectively reduce the dimensions of multi-dimensional spectral data and improve the accuracy of spectral detection of complex solution components, a method based on the gradient distribution of the multi-dimensional spectral image is proposed in this paper, aiming to extract the feature region, which is highly correlated with the solution concentration. In this research, firstly, photon propagation simulations in the wedge sample were performed using Monte Carlo, and the image of multi-dimensional spectra under different optical parameters was obtained. Furthermore, a method of extracting the position of the feature region based on the gradient distribution is proposed, according to the distribution characteristics of the image. Subsequently, a series of experiments were carried out of the visible-near-infrared spectra of 39 groups of mixed solutions, which composed of India Ink and Intralipid of sequential changed concentrations, to reveal the effectiveness of the feature region. The regression models for prediction of Intralipid concentration were developed by partial least square regression (PLSR). In contrast to the traditional spectral analysis method, the dimensionality of data was reduced to 1/9, by extracting the feature region of the multi-dimensional spectral data, while the average correlation coefficient (Rp) of prediction set reached 98.87%, and the average mean square error (RMSEP) of prediction set was 0.1502, showing an increase of 0.12% in Rp and a decrease of 6.12% in RMSEP, respectively. The results showed that models incorporating the feature region, improved the accuracy of component analysis of complex solutions with high-concentration and also the analysis speed and robustness of the model. A rapid, non-invasive, low-cost method is proposed in this paper, using visible-near-infrared spectroscopy with the wedge sample to analyze the components of complex solutions.

Bioelectronic Nose Based on Single-Stranded DNA and Single-Walled Carbon Nanotube to Identify a Major Plant Volatile Organic Compound (p-Ethylphenol) Released by Phytophthora Cactorum Infected Strawberries.

The metabolic activity in plants or fruits is associated with volatile organic compounds (VOCs), which can help identify the different diseases. P-ethylphenol has been demonstrated as one of the most important VOCs released by the Phytophthora cactorum (P. cactorum) infected strawberries. In this study, a bioelectronic nose based on a gas biosensor array and signal processing model was developed for the noninvasive diagnostics of the P. cactorum infected strawberries, which could overcome the limitations of the traditional spectral analysis methods. The gas biosensor array was fabricated using the single-wall carbon nanotubes (SWNTs) immobilized on the surface of field-effect transistor, and then non-covalently functionalized with different single-strand DNAs (ssDNA) through π–π interaction. The characteristics of ssDNA-SWNTs were investigated using scanning electron microscope, atomic force microscopy, Raman, UV spectroscopy, and electrical measurements, indicating that ssDNA-SWNTs revealed excellent stability and repeatability. By comparing the responses of different ssDNA-SWNTs, the sensitivity to P-ethylphenol was significantly higher for the s6DNA-SWNTs than other ssDNA-SWNTs, in which the limit of detection reached 0.13% saturated vapor of P-ethylphenol. However, s6DNA-SWNTs can still be interfered with by other VOCs emitted by the strawberries in the view of poor selectivity. The bioelectronic nose took advantage of the different sensitivities of different gas biosensors to different VOCs. To improve measure precision, all ssDNA-SWNTs as a gas biosensor array were applied to monitor the different VOCs released by the strawberries, and the detecting data were processed by neural network fitting (NNF) and Gaussian process regression (GPR) with high accuracy.

Applications of Fractional Lower Order Frequency Spectrum Technologies to Bearing Fault Analysis

The traditional spectral analysis method is used to study the characteristics of bearing fault signals in frequency domain, which is reasonable and effective in general cases. However, it is proved that the fault signals have heavy tails in this paper, which are α stable distribution, and 1 < α < 2, and even the noises belong to α stable distribution. Then the conventional spectral analysis methods degenerate and even fail under α stable distribution environment. Several improved frequency spectral analysis methods are proposed employing fractional lower order covariation or fractional lower order covariance in this paper, including fractional lower order Blackman‐Tukey covariation spectrum (FLOBTCS), fractional lower order periodogram covariation spectrum (FLOPCS), and fractional lower order welch covariation spectrum (FLOWCS). In order to suppress side lobe and improve resolution, we present novel fractional lower order autoregression (FLO‐AR) and fractional lower order autoregressive moving average (FLO‐ARMA) parameter model frequency spectrum methods, and the calculation steps are summarized. The proposed spectrum methods are compared with the existing methods based on second‐order statistics under Gaussian and SαS distribution environments, and the results show that the new algorithms have better performance than the traditional methods. Finally, the improved methods are applied to estimate frequency spectrums of the normal and outer race fault signals, and it is demonstrated that they are effective for fault diagnosis.

An Improved Spectral Analysis Method for Fatigue Damage Assessment of Details in Liquid Cargo Tanks

Errors will be caused in calculating the fatigue damages of details in liquid cargo tanks by using the traditional spectral analysis method which is based on linear system, for the nonlinear relationship between the dynamic stress and the ship acceleration. An improved spectral analysis method for the assessment of the fatigue damage in detail of a liquid cargo tank is proposed in this paper. Based on assumptions that the wave process can be simulated by summing the sinusoidal waves in different frequencies and the stress process can be simulated by summing the stress processes induced by these sinusoidal waves, the stress power spectral density (PSD) is calculated by expanding the stress processes induced by the sinusoidal waves into Fourier series and adding the amplitudes of each harmonic component with the same frequency. This analysis method can take the nonlinear relationship into consideration and the fatigue damage is then calculated based on the PSD of stress. Take an independent tank in an LNG carrier for example, the accuracy of the improved spectral analysis method is proved much better than that of the traditional spectral analysis method by comparing the calculated damage results with the results calculated by the time domain method. The proposed spectral analysis method is more accurate in calculating the fatigue damages in detail of ship liquid cargo tanks.

Hessian regularization by patch alignment framework

In recent years, semi-supervised learning has played a key part in large-scale image management, where usually only a few images are labeled. To address this problem, many representative works have been reported, including transductive SVM, universum SVM, co-training and graph-based methods. The prominent method is the patch alignment framework, which unifies the traditional spectral analysis methods. In this paper, we propose Hessian regression based on the patch alignment framework. In particular, we construct a Hessian using the patch alignment framework and apply it to regression problems. To the best of our knowledge, there is no report on Hessian construction from the patch alignment viewpoint. Compared with the traditional Laplacian regularization, Hessian can better match the data and then leverage the performance. To validate the effectiveness of the proposed method, we conduct human face recognition experiments on a celebrity face dataset. The experimental results demonstrate the superiority of the proposed solution in human face classification.

Two-Dimensional Continuous Wavelet Analysis and Its Application to Meteorological Data

AbstractThe two-dimensional continuous wavelet transform (2D CWT) has become an important tool to examine and diagnose nonstationary datasets on the plane. Compared with traditional spectral analysis methods, the 2D CWT provides localized spectral information of the analyzed dataset. It also has the advantage over the 2D discrete wavelet transform (DWT) in that it covers the domain of the analyzed data with a continuous analysis from which detailed, shift-invariant spectral information of different positions and orientations can be obtained. In this paper, a brief introduction of the 2D CWT and some of the most common wavelet mother functions are given, and some practical issues arising from the implementation and applications of the 2D CWT are discussed. The 2D CWT is applied to several test functions to illustrate the effects of the transforms. To demonstrate its practical application, the 2D CWT is used to analyze a set of meteorological data obtained from a numerical model stimulation.

Sine-fitting multiharmonic algorithms implemented by artificial neural networks

A new method designed to perform high-accuracy spectral analysis, based on ADALINE artificial neural networks (ANNs), is proposed. The proposed network is able to accurately calculate the fundamental frequency and the harmonic content of an input signal. The method is especially useful in high-precision digital measurement systems in which periodical signals are involved, i.e. digital watt meters. Most of these systems use spectral analysis algorithms as an intermediate step for the computation of the magnitudes of interest. The traditional spectral analysis methods require synchronous sampling, which introduce limitations to the sampling circuitry. Sine-fitting multiharmonics algorithms resolve the hardware limitations concerning the synchronous sampling but have some limitations with regard to the phase of the array of samples. The new implementation of sine-fitting multiharmonics algorithms based on ANN eliminates these limitations.

Comparative study of Hilbert–Huang transform, Fourier transform and wavelet transform in pavement profile analysis

This study employs the Hilbert–Huang transform (HHT), the wavelet transform and the Fourier transform to analyse the road surface profiles of three pavement profiles. The wavelet and Fourier transforms have been the traditional spectral analysis methods, but they are predicated on a priori selection of basis functions that are either of infinite length or have fixed finite widths. The central idea of HHT is the empirical mode decomposition, which decomposes a signal into basis functions called the intrinsic mode functions (IMFs). The Hilbert transform can then be applied to the IMFs to generate an energy–time–frequency spectrum called the Hilbert spectrum. The strength of HHT is the ability to process non-stationary and non-linear data. Unlike the Fourier transform, which transforms information from the time domain into the frequency domain, the HHT does not lose temporal information after transformation, i.e. energy–frequency information is maintained in the time domain. This paper attempts to reveal the frequency and energy content of the road profile data with the three methods mentioned as a means to establishing the most suitable way of characterising the pavement profiles in terms of ride quality. In performing the analyses, the nature and behaviour of the road profiles as indicated by the literature are taken into account, that road profiles are non-stationary and are inherently non-Gaussian. It is demonstrated that HHT offers more flexibility in terms of detailed profile analysis and description, which can be beneficial to pavement profile analysts in understanding the effects on vehicle vibrations and ride quality.

Ridge Distance Estimation in Fingerprint Images: Algorithm and Performance Evaluation

It is important to estimate the ridge distance accurately, an intrinsic texture property of a fingerprint image. Up to now, only several articles have touched directly upon ridge distance estimation. Little has been published providing detailed evaluation of methods for ridge distance estimation, in particular, the traditional spectral analysis method applied in the frequency field. In this paper, a novel method on nonoverlap blocks, called the statistical method, is presented to estimate the ridge distance. Direct estimation ratio (DER) and estimation accuracy (EA) are defined and used as parameters along with time consumption (TC) to evaluate performance of these two methods for ridge distance estimation. Based on comparison of performances of these two methods, a third hybrid method is developed to combine the merits of both methods. Experimental results indicate that DER is 44.7%, 63.8%, and 80.6%; EA is 84%, 93%, and 91%; and TC is 0.42, 0.31, and 0.34 seconds, with the spectral analysis method, statistical method, and hybrid method respectively.

Nonparametric Analysis Of Time-Resolved Fluorescence Data Based On The Laguerre Expansion Technique.

To estimate the intrinsic fluorescence intensity decay of a compound, the excitation light pulse must be deconvolved from the measured fluorescence pulse trace. The most commonly used deconvolution method is the multiexponential least-square iterative reconvolution (LSIR) technique. A variant of LSIR in which the intrinsic fluorescence intensity decay is expressed as an expansion on the discrete time Laguerre basis, was recently introduced. In this study, the performance of the Laguerre deconvolution technique was successfully tested with simulated and fluorescence standard data. It was also demonstrated that the Laguerre deconvolution presents a number of advantages over the classical multiexponential LSIR, including less expensive computational resolution, and the property to generate a unique set of expansion coefficients highly correlated with the intrinsic lifetimes. A novel method for concentration estimation based on the analysis of the Laguerre expansion coefficients was also proposed and successfully applied to different fluorescence standard mixtures, performing even better (error<2%) than more traditional methods of spectral analysis, such as PCR (error<7%) and PLS (error<10%). These findings suggest that the use of Laguerre expansion coefficients represents an alternative nonparametric approach to characterize and discriminate biological systems, in terms of their spectral and lifetime characteristics.

Cross sensitivity reduction of gas sensors using genetic neural network

The IR absorption method is a traditional spectral analysis method in the analysis of gas components. Nevertheless, the absorption spectrum lines of a certain kind of gas often intercross with those of another, which means that the absorption peak of one kind of gas is close to that of another. Thus, when these kinds of gases are mixed together, there will be cross-sensitivity in the concentration measurement using the gas sensor system. We adopt the genetic neural network algorithm to recognize the patterns of the mixed gas with two components and three components respectively in the simulation recognition. The staged-sectional recognition method is used to increase the recognition precision at various critical values of the mixed gases. The simulation results indicate that the staged-sectional method based on the genetic neural networks decreases the cross-sensitivity of the gas sensor and increases the recognition precision at various critical values.

Structural Similarity and IR Spectrum Modeling Using the Spectrum–Fragment Composition Database

It is shown on numerous examples that IR spectra of organic compounds may be modeled by using the spectra of the structural analogs selected from the database. This paper presents a technique for model construction and the results of modeling along with some examples of model and experimental spectra. The applicability of this approach to modern and traditional methods of spectral analysis is discussed.

Assessment of 3D proton MR echo-planar spectroscopic imaging using automated spectral analysis.

For many clinical applications of proton MR spectroscopic imaging (MRSI) of the brain, diagnostic assessment is limited by insufficient coverage provided by single- or multislice acquisition methods as well as by the use of volume preselection methods. Additionally, traditional spectral analysis methods may limit the operator to detailed analysis of only a few selected brain regions. It is therefore highly desirable to use a fully 3D approach, combined with spectral analysis procedures that enable automated assessment of 3D metabolite distributions over the whole brain. In this study, a 3D echo-planar MRSI technique has been implemented without volume preselection to provide sufficient spatial resolution with maximum coverage of the brain. Using MRSI acquisitions in normal subjects at 1.5T and a fully automated spectral analysis procedure, an assessment of the resultant spectral quality and the extent of viable data in human brain was carried out. The analysis found that 69% of brain voxels were obtained with acceptable spectral quality at TE = 135 ms, and 52% at TE = 25 ms. Most of the rejected voxels were located near the sinuses or temporal bones and demonstrated poor B0 homogeneity and additional regions were affected by stronger lipid contamination at TE = 25 ms.

AR and ARMA spectral estimation

Autoregressive (AR) and Autoregressive-moving average (ARMA) methods of spectral analysis have been developed and are being increasingly used as alternatives to traditional methods of spectral analysis. Two of these methods developed by Marple and Friedlander are tested in this study by using generated data from models with known spectra. The Blackman-Tukey spectral estimates are also compared to the Marple and Friedlander estimates. The variability of the Marple and Friedlander estimates with sample sizes is investigated. Although both Marple's and Friedlander's methods are satisfactory, Friedlander's method is preferred because of its ability to handle a wider class of models.