Spike detection approaches for noisy neuronal data: Assessment and comparison

Process Monitoring and Fault Detection using Empirical Mode Decomposition and Singular Spectrum Analysis

HighlightsWhat are the main findings?Time-frequency features extracted using EMD, DWT, and combined EMD+DWT effectively separate earthquake and noise signals.Random Forest classifier with Lasso-selected EMD+DWT features achieved 100% accuracy, specificity, and sensitivity.What are the implication of the main findings?Time-frequency-based feature extraction and selection improve real-time earth-quake detection.The approach provides a robust foundation for operational monitoring and ear-ly-warning systems.Earthquakes are sudden and destructive natural events caused by tectonic movements in the Earth’s crust. Although they cannot be predicted with certainty, rapid and reliable detection is essential to reduce loss of life and property. This study aims to automatically distinguish earthquake and noise signals from real seismic data by analyzing time-frequency features. Signals were scaled using z-score normalization, and extracted with Empirical Mode Decomposition (EMD), Discrete Wavelet Transform (DWT), and combined EMD+DWT methods. Feature selection methods such as Lasso, ReliefF, and Student’s t-test were applied to identify the most discriminative features. Classification was performed with Ensemble Bagged Trees, Decision Trees, Random Forest, k-Nearest Neighbors (k-NN), and Support Vector Machines (SVM). The highest performance was achieved using the RF classifier with the Lasso-based EMD+DWT feature set, reaching 100% accuracy, specificity, and sensitivity. Overall, DWT and EMD+DWT features yielded higher performance than EMD alone. While k-NN and SVM were less effective, tree-based methods achieved superior results. Moreover, Lasso and ReliefF outperformed Student’s t-test. These findings show that time-frequency-based features are crucial for separating earthquake signals from noise and provide a basis for improving real-time detection. The study contributes to the academic literature and holds significant potential for integration into early warning and earthquake monitoring systems.

[Formula: see text]Si Nuclear Magnetic Resonance (NMR) can measure the molecular structure of silicate in oilfield reinjection water. However, noise in [Formula: see text]Si NMR spectra (NMRS) affects the determination of silicate molecular structure type. To solve this problem, a new peak fitting method (Two-step Greedy-Singular Spectrum Analysis-Gaussian Fitting Method, TSG-SSA-GFM) is proposed in this paper. This method first uses TSG to determine the embedding dimension, then uses SSA to determine the characteristic peak position. Finally, GFM is used to calculate the molar ratio of characteristic peaks. The results show that TSG can quickly determine the embedding dimension and reduce computation by at least 50% vs. the global ergodic method. The mean deviation of characteristic peak positions determined by SSA is 0.07 ppm, while Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD) cannot determine characteristic peaks of [Formula: see text]Si NMRS containing overlapping peak. The average [Formula: see text]-squared of Gaussian fitting of [Formula: see text]Si NMRS is 98.4% while Lorentzian is 90.6%. Therefore, this study provides an important method for quantitative analysis of [Formula: see text]Si NMRS.

A brainwave classification, which does not involve any limb movement and stimulus for character-writing applications, benefits impaired people, in terms of practical communication, because it allows users to command a device/computer directly via electroencephalogram signals. In this paper, we propose a new framework based on Empirical Mode Decomposition (EMD) features along with the Gaussian Mixture Model (GMM) and Kernel Extreme Learning Machine (KELM)-based classifiers. For this purpose, firstly, we introduce EMD to decompose EEG signals into Intrinsic Mode Functions (IMFs), which actually are used as the input features of the brainwave classification for the character-writing application. We hypothesize that EMD along with the appropriate IMF is quite powerful for the brainwave classification, in terms of character applications, because of the wavelet-like decomposition without any down sampling process. Secondly, by getting motivated with shallow learning classifiers, we can provide promising performance for the classification of binary classes, GMM and KELM, which are applied for the learning of features along with the brainwave classification. Lastly, we propose a new method by combining GMM and KELM to fuse the merits of different classifiers. Moreover, the proposed methods are validated by using the volunteer-independent 5-fold cross-validation and accuracy as a standard measurement. The experimental results showed that EMD with the proper IMF achieved better results than the conventional discrete wavelet transform (DWT) feature. Moreover, we found that the EMD feature along with the GMM/KELM-based classifier provides the average accuracy of 77.40% and 80.10%, respectively, which could perform better than the conventional methods where we use DWT along with the artificial neural network classifier in order to get the average accuracy of 80.60%. Furthermore, we obtained the improved performance by combining GMM and KELM, i.e., average accuracy of 80.60%. These outcomes exhibit the usefulness of the EMD feature combining with GMM and KELM based classifiers for the brainwave classification in terms of the Character-Writing application, which do not require any limb movement and stimulus.

Aerodynamic instabilities in centrifugal compressors are dangerous phenomena affecting machines efficiency and in severe cases leading to failures of the compressing system due to high amplitude vibrations. Efficient instabilities detection during compressor operation is a challenge of utmost importance from economical and safety point of view. The most sensitive detection is possible with use of a pressure signal from inside of the compressor because specific pressure patterns are the first symptoms of instabilities. Detection of aerodynamic instabilities results in specific challenges, as the pressure signal is often very noisy and contains high amount of randomness. Surge - most severe instability, can develop very quickly. Therefore, the method of detection should be sensitive but also robust and quick. Another common instability, inlet recirculation is less dangerous, but it results in decrease of efficiency, which is to be avoided. Inlet recirculation often happens before surge, thus its presence can be used for surge proximity detection. The aim of this study is to investigate and compare the performance of two non-linear processing methods - Empirical Mode Decomposition (EMD) and Singular Spectrum Analysis (SSA) in the context of aerodynamic instabilities detection - inlet recirculation and surge. The comparison focuses on the robustness, sensitivity and pace of detection - crucial parameters for a successful detection method. It is shown that both methods perform similarly within the analyzed bounds for both instabilities. A slight advantage of SSA may be noticed for surge due to lower dispersion of the indicator value for the same conditions.

Unobtrusive cardiorespiratory monitoring in the period of sleep could facilitate researchers for sleep analysis and sleep disorders diagnosis. With the recent advancement of biomedical health informatics, it is possible to monitor individual's sleep disorders through cardiovascular parameters. This paper proposes a polyvinylidene fluoride (PVDF) sensor based system for unobtrusive measurements of respiration and heartbeat while sleeping; the sensing unit is deployed on the mattress below the back area of the subject. The PVDF sensor generates the sensory activations due to pulsatile pressure change caused by respiratory and heartbeat movement. To extract and separate the heartbeat and respiration accurately, the method singular spectrum analysis (SSA) was applied while the empirical mode decomposition (EMD) and wavelet analysis methods were used for comparison. Experiments are conducted on two healthy subjects to obtain the respiratory rate and heart rate. To verify and compare the performance of the proposed system, the polysomnography (PSG) system was used to record the thorax effort signal and electrocardiography (ECG) as the gold standard for extracting the respiratory rate and heart rate respectively. The average error rate (AER) of respiratory rate extraction is noted 8.53%, with segmenting the data as 2 mins epochs. Meanwhile, the AERs of the heart rate are 8.01%, 8.80%, 8.55% by utilizing the SSA, EMD and wavelet methods respectively. Overall, the proposed system can obtain accurate respiration and heart rate detection. And the novelty lies in introducing a flexible PVDF sensor which can be combined with the sleep mat for cardiorespiratory signal monitoring in home environment; and new SSA algorithm to extract respiration and heart rates. The experimental results are promising which demonstrated that the proposed system satisfied the basic requirements for cardiorespiratory signal monitoring during sleep.

This paper presents a new approach for human automatical detection based on ultrasound signal processing. Two signal decomposition techniques, empirical mode decomposition and discrete wavelet transform are used in the presented method. Performance comparison of these two decomposition methods when applied to this specific ultrasound signal is given. Described method includes ultrasonic signal decomposition to enhance signal specific features and increase signal to noise ratio with the following decision rules based on adaptive thresholding. The resulting detection performance of the proposed method using empirical mode decomposition was better to compare to discrete wavelet transform and resulted in 70% correctly identified healthy and 82%, 97% and 100% correctly identified cataract cases in the incipience, immature and mature subject groups, respectively. Discussion is given on the reasons of different results and the differences between the two used signal decomposition techniques.

Empirical mode decomposition (EMD) is particularly useful in analyzing nonstationary and nonlinear time series. However, only partial data within boundaries are available because of the bounded support of the underlying time series. Consequently, the application of EMD to finite time series data results in large biases at the edges by increasing the bias and creating artificial wiggles. This study introduces a new two-stage method to automatically decrease the boundary effects present in EMD. At the first stage, local polynomial quantile regression (LLQ) is applied to provide an efficient description of the corrupted and noisy data. The remaining series is assumed to be hidden in the residuals. Hence, EMD is applied to the residuals at the second stage. The final estimate is the summation of the fitting estimates from LLQ and EMD. Simulation was conducted to assess the practical performance of the proposed method. Results show that the proposed method is superior to classical EMD.

In this study, normal (healthy), pre-seizure and seizure states were analyzed using EEG data from BONN database. In this paper, we propose a new method of using SVM, KNN and Decision Tree for classification analysis in order to improve the detection accuracy of seizure. First, due to the presence of significant noise in the EEG signals, for signal pre-processing we performed noise removal. Second, two methods of frequency space transformation were used such as Discrete Wavelet Transform (DWT), separating signal into sub-bands and Empirical Mode Decomposition (EMD) technique to decompose signal into the Intrinsic Mode Functions (IMFs). Before the classification, the statistical moments of the signals in the frequency domain were obtained for feature extraction. Using these features, the performances of Support Vector Machines (SVM), Decision Tree, K-nearest Neighborhood (KNN) classifiers were analyzed. The experiment results show that the most accurate detection of epilepsy was obtained by applying EMD method with classifiers SVM, KNN and Decision Tree, and such algorithm with EMD can achieve accuracy for normal, pre-seizure, and seizure equal to 100%, 100% and 96.67%, respectively.

Haemorrhages (HAs) presence in fundus images is one of the most important indicators of diabetic retinopathy that causes blindness. In this regard, accurate grading of HAs in fundus images is crucial for appropriate medical treatment. The purpose of this Letter is to assess the relative performance of statistical features obtained with three different multi-resolution analysis (MRA) techniques and fed to support vector machine in grading retinal HAs. Considered MRA techniques are the common discrete wavelet transform (DWT), empirical mode decomposition (EMD), and variational mode decomposition (VMD). The obtained experimental results show that statistical features obtained by EMD, VMD, and DWT, respectively, achieved 88.31% ± 0.0832, 71% ± 0.1782, and 64% ± 0.0949 accuracies. It also outperformed VMD and DWT in terms of sensitivity and specificity. Thus, the EMD-based features are promising for grading retinal HAs.

The interpretation of electrochemical noise (EN) data has long been under discussion. Throughout the years, many data analysis techniques have been proposed for this purpose. As a starting point, procedures and parameters that enable identification of, or discrimination between, general and localized corrosion processes through EN are critically discussed. It is important to consider which type of information is required for a specific application. EN can be used to determine the barrier properties of a protective coating, and can therefore provide quantitative information about corrosion processes, although its most interesting ability is to provide information on corrosion characteristics. EN signals consist of a direct current (DC) component, or trend, with superimposed fluctuations. In electrochemical potential noise (EPN), this DC component represents the open corrosion potential (OCP) of the system under study. In electrochemical current noise (ECN), the DC component can e.g. be generated by electrode asymmetry. Any DC drift should be carefully removed prior to further data analysis. A procedure is introduced to accurately define the DC drift in EN signals using either of two time-frequency data analysis techniques: discrete wavelet transform (DWT) or empirical mode decomposition (EMD). Consistent and reliable information can be obtained from EN data when a data analysis procedure is selected that on the one hand has a high discrimination ability and on the other hand yields a descriptive parameter that is directly associated to the underlying physico- chemical process. Preferably, the information is obtained without the need for subjective a-priori limitations or assumptions concerning the nature of the process under investigation. These requirements are met by the Hilbert-Huang transform, which is based on the EMD. The result is a Hilbert spectrum in which local frequency information, so-called instantaneous frequencies, of EN signals, is visualized. The use of Hilbert spectra for the characterization of EN data in corrosion studies is introduced, based on the (general and localized) corrosion characteristics of carbon steel and stainless steel AISI304. A highly detailed decomposition of the original ECN and EPN data is provided in time and frequency simultaneously. This allows distinguishing between different corrosion characteristics based on their EN signals. Hilbert spectra also provide the possibility to analyze only transients present in the EN signals that originate from localized corrosion processes on AISI304. Initial identification of transients is based on the transient shape. Analysis of instantaneous frequency information present in these transients enables improved differentiation between corrosion characteristics as compared to data analysis using Hilbert spectra or energy distribution plots (determined from DWT) without transient analysis. The applicability of transient analysis through Hilbert spectra of ECN signals is further investigated for Ce-based inhibition of aluminium alloy AA2024-T3 and for detection and identification of microbiologically influenced corrosion (MIC). Transient analysis allows detection of changes in corrosion characteristics, i.e. the evolution of corrosion inhibition of AA2024-T3 by Ce-ions, with time. The initial procedure of transient detection is further developed, comprising of automatic detection of specific areas of interest in Hilbert spectra between 10E-1 Hz and 1 Hz, corresponding with the occurrence of transients in the respective ECN signals. Regarding the detection of MIC, together with monitoring of the OCP and microscopic observations, the development of ECN transients generated by localized corrosion processes could be attributed to the presence and activity of sulphate-reducing bacteria. These transients are related to the existence of pits in the carbon steel surface, underneath the attached biofilm. Finally, practical aspects and configurations for electrochemical noise measurements (ENM) are discussed. ENM can be applied in a hand-held solution, or for permanent monitoring. Analogous to the selection of the appropriate data analysis procedure to obtain the information of interest, it is important to consider the required application to select the most suitable configuration. ENM is a potentially interesting technique because of its non-intrusive nature, the robust sensor configurations, the ability to identify localized corrosion processes and ease of use. The most complicated aspect of ENM is the interpretation of the EN signals, for which the data analysis procedure can be fully automated if required.

The electrocardiographic (ECG) signal is a major artifact during recording the surface electromyography (SEMG). Removal of this artifact is one of the important tasks before SEMG analysis for biomedical goals. In this paper, the application of discrete wavelet transform (DWT) and empirical mode decomposition (EMD) for elimination of ECG artifact from SEMG is investigated. The focus of this research is to reach the optimized number of decomposed levels using mean power frequency (MPF) by both techniques. In order to implement the proposed methods, ten simulated and three real ECG contaminated SEMG signals have been tested. Signal-to-noise ratio (SNR) and mean square error (MSE) between the filtered and the pure signals are applied as the performance indexes of this research. The obtained results suggest both techniques could remove ECG artifact from SEMG signals fair enough, however, DWT performs much better and faster in real data.

Most often, climate change signals are slow moving (in human terms), low amplitude changes embedded in high amplitude, noisy data. There are many techniques for extracting such signals. This paper introduces a technique which in contrast with empirical methods produces a decomposition of a time series into a set of equations plus a This is referred to as stepwise symbolic secomposition (SSD). The extraction of symbolic equations in lieu of empirical functions assists with characterization of time series. This paper takes two examples of climate data, and applies two different techniques for characterising the low amplitude, slow change embedded therein. The record of CO2 levels at Mauna-Loa since March 1958, is used to demonstrate and contrast empirical mode decomposition (EMD), and SSD. The mean monthly tidal gauge records from the small number of gauges which have more than 120 years of data are analysed in more detail by SSD, then three techniques are used to characterize the residual. The techniques are (a) LOESS smoothing, (b) EMD, and (c) high order polynomial regression. SSD uses a genetic programming system called Eureqa from Cornell Creative Machines Lab guided by an information metric, to extract the most compact informative function it can at each step; the process is repeated on the residuals until no sufficiently informative function is found. EMD and SSD are in stark contrast in the order in which signals are decomposed. EMD extracts high frequency components first. SSD extracts components based on a mixture of parsimonious representation and variance explained. EMD leaves a low frequency filtrate of the signal in its residue, SSD tends to operate as a broad band filter, leaving high frequency noise plus a possible low frequency signal. EMD, LOESS and polynomial fitting all serve to extract low frequency components of the signal from this SSD residual. For the set of tidal gauge data, in the absence of a change in global sea level rise, the SSD procedure should randomize all segments of residual signals equally - the residual should be whitened with respect to the initial signal. It is shown however that late 20th century portions of the residuals show behaviour that is consistent with accelerating sea level rise in keeping with the bulk of the literature.

Comparative analysis of ANN performance of four feature extraction methods used in the detection of epileptic seizures

In power system engineering, Transmission Lines (TLs) plays a major role, an important goal of this is to transmit the generated electricity to various distribution units. Transmission lines act as a bridge between consumers and the powerhouse. In the transmission lines, the faults often occur because they are located in open space. Quick detection and recovery of transmission lines faults are necessary to maintain the normal operation of the power system. To overcome these disturbances, techniques like Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD) are used in fault detection and diagnosis. DWT is applied to decompose the fault transients, because it can simultaneously extract the data, in both time and frequency domain. EMD is used to decompose the transmission line voltages into Intrinsic Mode Function (IMFs). Four types of fault signals are generated from the grid-connected power system. Transmission line faults are created using MATLAB/Simulink.