Blind Source Separation Research Articles

Non-targeted volatilomics for the authentication of saffron by gas chromatography-ion mobility spectrometry and multivariate curve resolution.

In the present contribution, a novel non-targeted volatilomic study based on headspace GC-IMS (HS-GC-IMS) was developed for the authentication and geographical origin discrimination of saffron. In this regard, multivariate curve resolution-alternating least squares (MCR-ALS) was employed to recover the pure GC elution and IMS profiles of saffron metabolites. Iranian saffron samples from seven important areas were analyzed by HS-GC-IMS. The resulting second-order GC-IMS datasets were organized in a augmented matrix and processed using MCR-ALS with various constraints. The MCR-ALS resolved GC profiles were analyzed by different pattern recognition techniques; principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA) and data driven-soft independent modeling of class analogy (DD-SIMCA). The saffron samples were assigned to their seven geographical origins with an accuracy of 89.0%. Additionally, four adulterants (style, safflower, madder and calendula) were reliably detected with over 94.0% accuracy. In this context, GC-IMS substantially outperformed the commonly used FT-NIR spectroscopy approach.

Predicting carbon dioxide emissions using deep learning and Ninja metaheuristic optimization algorithm

This paper provides a novel approach to estimating CO₂ emissions with high precision using machine learning based on DPRNNs with NiOA. The data preparation used in the present methodology involves sophisticated stages such as Principal Component Analysis (PCA) as well as Blind Source Separation (BSS) to reduce noise as well as to improve feature selection. This purified input dataset is used in the DPRNNs model, where both short and long-term temporal dependencies in the data are captured well. NiOA is utilized to tune those parameters; as a result, the prediction accuracy is quite spectacular. Experimental results also demonstrate that the proposed NiOA-DPRNNs framework gets the highest value of R2 (0.9736), lowest error rates and fitness values than other existing models and optimization methods. From the Wilcoxon and ANOVA analyses, one can approve the specificity and consistency of the findings. Liebert and Ruple firmly rethink this rather simple output as a robust theoretic and empirical framework for evaluating and projecting CO2 emissions; they also view it as a helpful guide for policymakers fighting global warming. Further study can build up this theory to include other greenhouse gases and create methods enabling instantaneous tracking for sophisticated and responsive approaches.

A U-Net based partial convolutional time-domain separation model to identify motor units from surface electromyographic signals in real time.

This study proposed a U-Net based partial convolutional time-domain model for a real-time high-density surface electromyography (HD-sEMG) decomposition. The model combines U-Net and a separation block containing partial convolution, aiming to efficiently identify motor units (MUs) without preprocessing. The proposed U-Net based network was trained by the HD-sEMG signals with innervation pulse trains (IPTs) labels, and the results are compared between different step sizes, noises, and model structures under the sliding time window with 120 sampling points. The U-Net based model got an accuracy greater than 94% under simulated signals and 85% under experimental signals, and identified more MUs than the structures based on convolutional neural network (CNN) and temporal convolutional network (TCN). The average latency of the U-Net based model is only 64ms (a window duration time plus the prediction time) under the step size 20 data in both types of signals, and can be generalized to new data at different signal-to-noise (SNR). The efficiency of the proposed model is significantly higher than traditional methods such as gCKC. Meanwhile, the accuracy of the proposed model was not significantly different from the gCKC. In addition, the performance of the network under different step sizes of the sliding time window was verified. The experimental results indicate that the U-Net based model provides an efficient framework for blind source separation (BSS) of EMG signals, expanding the application of EMG signals in neural interaction.

Application of signal separation to diffraction image compression and serial crystallography

We present here a methodology for real-time analysis of diffraction images acquired at a high frame rate (925 Hz) and its application to macromolecular serial crystallography at ESRF. We introduce a new signal-separation algorithm, able to distinguish the amorphous (or powder diffraction) component from the diffraction signal originating from single crystals. It relies on the ability to work efficiently in azimuthal space and is implemented in pyFAI, the fast azimuthal integration library. Two applications are built upon this separation algorithm: a lossy compression algorithm and a peak-picking algorithm. The performances of both are assessed by comparing data quality after reduction with XDS and CrystFEL.

A modified blind source separation algorithm for underdetermined structural modal analysis

A modified blind source separation algorithm for underdetermined structural modal analysis

Investigation of charged Higgs bosons production from vectorlike T quark decays at eγ collider

Within the extended framework of the two-Higgs-doublet model type II (2HDM-II), enhanced by a vectorlike quark (VLQ) doublet TB, we present a comprehensive analysis of the process e−γ→bνeT¯ at future high-energy eγ colliders, focusing on the decays T¯→H−b¯ and H−→t¯b. Using current theoretical and experimental constraints, we calculate production cross sections for both unpolarized and polarized beams at center-of-mass energies of s=2 and 3 TeV, demonstrating that polarized beams significantly enhance detection prospects by increasing production rates. By analyzing kinematic distributions, we establish optimized selection criteria to effectively separate signal events from background. At s=2 TeV with an integrated luminosity of 1500 fb−1, we find exclusion regions within sRd∈[0.085,0.16] for mT∈[1000,1260] GeV and a discovery potential within sRd∈[0.14,0.17] for mT∈[1000,1100] GeV, with these regions expanding to sRd∈[0.05,0.15] for mT∈[1000,1340] GeV and sRd∈[0.11,0.17] for mT∈[1000,1160] GeV at 3000 fb−1. At s=3 TeV and 1500 fb−1, we identify exclusion regions of sRd∈[0.055,0.135] for mT∈[1000,1640] GeV and discovery regions of sRd∈[0.09,0.15] for mT∈[1000,1400] GeV, which further expand to sRd∈[0.028,0.12] for mT∈[1000,1970] GeV and sRd∈[0.04,0.122] for mT∈[1000,1760] GeV at 3000 fb−1. Our findings emphasize the increased detection potential at higher center-of-mass energies, particularly at 3 TeV compared to 2 TeV, with notable improvements when polarized beams are utilized. We also account for the effects of initial state radiation, beamstrahlung, and systematic uncertainties, which influence both exclusion and discovery prospects. Published by the American Physical Society 2025

Leveraging overfitting for good - Two-Step Deep Image Prior Model For Seismic Denoising

Accurate separation of signal and noise constitutes a fundamental prerequisite for achieving high-resolution seismic imaging. A notable recent advancement in this domain is the Deep Image Prior (DIP), an unsupervised deep learning method leveraging deep neural networks (DNNs). The success of this approach lies in the adoption of autoencoders that enables the adaptive extraction of high-fidelity data features. However, establishing an optimal balance between noise suppression and signal preservation remains a non-trivial challenge for DIP-based seismic denoising methods, which is affected by the potential issue of overfitting. This arises from the inappropriate selection of a network architecture and the corresponding hyperparameters, especially the number of training epochs, which strongly influence the learning capacity and feature extraction capabilities of the model. In response to this challenge, we introduce Two-Step DIP (TSDIP), a novel denoising method that exploits overfitting to enhance seismic data quality. In the initial stage, the proposed DNNs are intentionally trained to overfit by effectively attenuating high-frequency noise from the input data. Subsequently, the proposed DNNs are employed iteratively to suppress any residual noise in the newly processed data without damaging useful signal. The overfitting in the first step helps precondition the data to be at a lower noise level while preserving as much as fine-scale features in the signal. To employ an optimal balance, we carefully determine an ideal number of epochs, which is consistently applied in both denoising steps. To assess the effectiveness of the TSDIP method, we present the test results derived from 3D synthetic and field seismic datasets. Our analysis indicates that TSDIP effectively reduces strong noise while preserving key seismic details through the use of overfitting.

Underdetermined blind source separation with adaptive chirp mode decomposition for compound rolling bearing fault signals

Underdetermined blind source separation with adaptive chirp mode decomposition for compound rolling bearing fault signals

Tracking solid electrolyte interphase dynamics using operando fibre-optic infra-red spectroscopy and multivariate curve regression

As batteries drive the transition to electrified transportation and energy systems, ensuring their quality, reliability, lifetime, and safety is crucial. While the solid electrolyte interphase (SEI) is known to govern these performance characteristics, its dynamic nature makes understanding its nucleation, growth, and composition an ambitious, yet elusive aspiration. This work employs chalcogenide fibres embedded in negative electrode materials for operando Infra-red Fibre-optic Evanescent Wave Spectroscopy (IR-FEWS), combined with Multivariate Curve Resolution by Alternating Least Squares (MCR-ALS) algorithms for spectra analysis. By establishing molecular fingerprints that can be used to identify reaction products, IR-FEWS combined with MCR-ALS enables improved understanding of SEI evolution during cell formation with notable differences stemming from electrolyte or anode material. For example, despite operating at an elevated potential, lithium titanate’s SEI has intrinsic instability, evidenced by continued carbonate formation. This approach leads the hunt for the SEI down a new path, giving empirical formulations theoretical roots.

Hybrid Bilayer Interfaces within Reversed-Phase Chromatographic Silica Formed by Self-Assembly of Long-Chain Primary Alcohols.

Modification of silica interfaces by covalent attachment of functional ligands is a primary means of controlling the interfacial chemistry of porous silicas used in separations, environmental cleanup, and biosensing. Recently, modification of hydrophobic, n-alkyl-silane-functionalized interfaces has been achieved through self-assembly of zwitterionic phospholipids or mixed-charged surfactants to form "hybrid bilayers", producing interfaces that mimic lipid-bilayer partitioning and provide shape-selective partitioning of aromatic hydrocarbons. Charged headgroups, however, introduce electrostatic interactions that strongly influence the retention of ionizable solutes and require careful control over pH and ionic strength in the solution phase. In this work, we propose modification of C18-functionalized chromatographic silica surfaces through self-assembly of long-chain primary alcohols to form uncharged hybrid-bilayer surfaces. Hybrid bilayers formed from alcohols ranging from C12OH to C22OH are investigated with in situ confocal-Raman microscopy, and the spectra indicate that they form highly ordered n-alkane structures, with order increasing as a function of alcohol chain length. Temperature-dependent Raman spectra of C12OH-C22OH hybrid bilayers were collected to investigate their melting transitions. Multivariate curve resolution of these spectra show broad, two-component melting transitions, indicating alcohol and C18 alkyl chains melt simultaneously. These results suggest an interdigitated interfacial structure, where the hydrocarbon chains of the adsorbed alcohol extend into the underlying C18 chains, ordering both layers. Interdigitation is confirmed by a temperature-dependent study of a deuterated C16-OH bilayer, where spectrally resolved Raman bands from deuterated and protiated hydrocarbons melt together. Finally, n-alkyl alcohol bilayers were tested for protein repellency, where no protein adsorption was observed when equilibrated with ∼1 mg/mL bovine serum albumin. Bilayers C16OH in chain length are shelf stable at refrigerated temperatures for months. These results demonstrate long-chain alcohol bilayers can be utilized to control the interfacial hydrocarbon structure of C18-modified silica and have potential for use in separations, biosensing, and anti-biofouling applications.

Array Coil Design and Experimental Verification for Separation of Tower Grounding Pulsed Eddy Current Excitation and Response Magnetic Field Signals

Array Coil Design and Experimental Verification for Separation of Tower Grounding Pulsed Eddy Current Excitation and Response Magnetic Field Signals

Coordinated regulation of eosinophil production and migration by glucocorticoids, prostaglandins, and cysteinyl-leukotrienes.

The spectrum of eosinophil functions has expanded from fighting helminths to multiple novel roles in malignancy, infection, cancer, and metabolism. In asthma, glucocorticoids, prostaglandins (PG), and cysteinyl-leukotrienes (LT) regulate eosinophil biology through separate signaling pathways. Here we've evaluated the complex interplay between Dexa, PGE2, and CysLTs in eosinopoiesis and eosinophil biology in an allergic asthma model. We used different inbred mouse strains to probe the interactions between these agents in eosinophil differentiation and maturation in bone marrow culture. Flow cytometry and histological analyses evaluated eosinophil precursor proliferation, maturation, and VLA-4 expression. The in vivo function of eosinophils was assessed by their in vivo migration into allergen-challenged sites. Eosinophil production in IL-5-stimulated bone marrow cultures is enhanced by dexamethasone but suppressed by PGE2, which triggers eosinophil apoptosis via inducible NO synthase (iNOS). Dexamethasone-primed cultures contain mostly immature eosinophils; by contrast, dexamethasone associated with PGE2 leads to the production of mature eosinophils (without inducing eosinophil apoptosis), by a mechanism independent of iNOS. Interaction between dexamethasone and LTD4 in culture produces mostly mature eosinophils expressing VLA-4, capable of migration into the lungs in ovalbumin-sensitized and -challenged mice. The combination of dexamethasone and either PGE2 or LTD4 - both mediators of allergic inflammation - supports the maturation of eosinophils (overcoming the maturation blockade observed with dexamethasone alone), which are functional in an in vivo model of asthma.

Introducing the "Cockpit Party Problem": Blind Source Separation Enhances Aircraft Cockpit Speech Transcription

Cockpit voice recorders (CVRs) are one of the two mandatory flight recording devices embarked in commercial aircraft. Its analysis is crucial to understand the context of an air incident or accident. However, in such scenarios, when the audio recordings are usable, CVR may contain strong mixtures of crew member speech signals, radio communications, and cockpit alarms. However, contrary to the “cocktail party problem” that blind source separation (BSS) aims to tackle, modeling CVR mixtures—that are here named the “cockpit party problem”—was never done before. In this paper, the authors thus propose a CVR mixture model and highlight its limitations. While not trivial—even in a two-source scenario—BSS methods can be applied to real CVR recordings. It is found that taking into account several BSS outputs provided by various methods may help audio analysts to transcribe the CVR data. That is, near 90% of unintelligible words can be transcribed from CVR recordings processed by BSS methods.

Estimation of Forest Canopy Height from Spaceborne Full-Waveform LiDAR Data Using a Bisection Approximation Decomposition Method

Forest canopy height (FCH) is a vital indicator for assessing forest health and ecosystem service capacity. Over the past two decades, full-waveform (FW) LiDAR has been widely employed for estimating forest biophysical variables due to its high precision in measuring vertical forest structures. However, the impact of terrain undulations on forest parameter estimation remains challenging. To address this issue, this study proposes a bisection approximation decomposition (BAD) method for processing GEDI L1B data and FCH estimation. The BAD method analyzes the energy composition of simplified echo signals and determines the fitting parameters by integrating overall signal energy, the differences in unresolved signals, and the similarity of inter-forest signal characteristics. FCH is subsequently estimated based on waveform peak positions. By dynamically adjusting segmentation points and Gaussian fitting parameters, the BAD method achieved precise separation of mixed canopy and ground signals, substantially enhancing the physical realism and applicability of decomposition results. The effectiveness and robustness of the BAD method for FCH estimation were evaluated using 2049 footprints across varying slope conditions in the Harvard Forest region of Petersham, Massachusetts. The results demonstrated that digital terrain models (DTMs) extracted using the GEDI data and the BAD method exhibited high consistency with the DTMs derived using airborne laser scanning (ALS) data (coefficient of determination R2 > 0.99). Compared with traditional Gaussian decomposition (GD), wavelet decomposition (WD), and deconvolution decomposition (DD) methods, the BAD method showed significant advantages in FCH estimation, achieved the smallest relative root mean square error (rRMSE) of 17.19% and greatest mean estimation accuracy of 84.57%, and reduced the rRMSE by 10.74%, 21.49%, and 28.93% compared to GD, WD, and DD methods, respectively. Moreover, the BAD method exhibited a significantly stronger correlation with ALS-derived canopy height mode data than the relative height metrics from GEDI L2A products (r = 0.84, p < 0.01). The robustness and adaptability of the BAD method to complex terrain conditions provide great potential for forest parameters using GEDI data.

Identification of operational modal parameters for a large floodgate under hydraulic excitation using the blind source separation improved by Hankel matrix singular value and adaptive variational mode decomposition

Identification of operational modal parameters for a large floodgate under hydraulic excitation using the blind source separation improved by Hankel matrix singular value and adaptive variational mode decomposition

A fast region of interest algorithm for efficient data compression and improved peak detection in high-resolution mass spectrometry.

Liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS) is commonly used for identification of compounds in complex samples due to the high chromatographic and mass spectral resolution provided. In subsequent data processing workflows, it is imperative to preserve this resolution to fully exploit the data. "Region of interest" (ROI) algorithms were introduced as a better alternative to equidistant binning for compressing HRMS data because they better preserve the mass spectral resolution. In this paper, we present a new ROI algorithm that improves on the selection of contiguous m/z traces, amongst others by introducing the concept of chromatographic filter, allows for an automated approach to optimise the admissible mass-to-charge deviation (δm/z) and can be used to match ROIs across multiple samples. The algorithm was tested on a LC-HRMS dataset comprised of 21 replicate injections of a wastewater effluent extract and assessed on its ability to correctly retrieve the ROI's relative to 57 compounds and match them across all injections. In summary, it achieved a ten-fold compression rate in on-disk storage at a noise threshold of 200 counts, and the median ROI length matched the observed chromatographic peak width (12-23 points). Correct ROI matching with a mass accuracy of 9ppm was observed for 52 compounds across all 21 injections with only one compound split between two adjacent m/z traces in six runs. Overall, the new algorithm performed favourably compared to the ROI algorithm currently used in the well-established ROI-MCR (multivariate curve resolution) workflow for deconvolution of HRMS chromatographic data.

Internal structural changes in age-related curved hair due to cyclical extension: An IR spectroscopy study.

This study aimed to identify structural changes in age-related curved hair (referred to as "YUGAMI" hair in Japanese) induced by cyclical extension using infrared (IR) spectroscopy coupled with chemometrics, such as multivariate curve resolution (MCR) and two-dimensional correlation spectroscopy (2DCOS). The hair fibres were stretched at a strain level of 0.3-N, and this operation was counted as one cycle and was repeated 500 cycles. Every 25 cycles, IR spectra were obtained using an attenuated total reflection IR (ATR-IR) microscope. The spectra were analysed by MCR to investigate the structural changes of keratin protein inside of the hair. 2DCOS was conducted to investigate the genesis pathway of SO3H in the 1000-1150 cm-1 range and the change in the secondary structure of keratin protein of hair in the 1600-1700 cm-1 range. For age-related curved hair, 0.3-N strain levels increased the CH2 and C=O signals after 125 cycles and the SO3H signal after 225 cycles, but no architectural modifications were observed in straight hair. Two results were obtained from 2DCOS. One was cyclical extension increased the S-SO3H signal, followed by SO3H signal. The other in the 1600-1700 cm-1 range indicated an increase in the secondary structural motif and β-structure after the formation of SO3H. This study shows that the process of internal transformations can be split into three steps in age-related curved hair caused by cyclical extension. The first step is the surface transfer of internal lipids, and the second step is two distinct patterns of disulfide bond cleavage: C-S and S-S fission. The final step is further transformation of the unravelled α-helices into β-structures which occurs above roughly 225 stretched. Our findings can be an effective factor in care the increase of curvature by cyclical extension such as everyday grooming.

Signal Processing Workflow for Suspect Screening in LC × LC-HRMS: Efficient Extraction of Pure Mass Spectra for Identification of Suspects in Complex Samples Using a Mass Filtering Algorithm.

The data processing workflows for comprehensive two-dimensional liquid chromatography (LC × LC) hyphenated to high-resolution mass spectrometry (HRMS) operated in data-independent acquisition (DIA) are limited compared to their one-dimensional counterparts. A two-step workflow is proposed to extract pure mass spectra from LC × LC-HRMS. First, a mass filtering (MF) algorithm groups ions belonging to the same compound based on their elution profile similarity in the first (1D) and second dimension (2D). Second, the filtered data are deconvoluted using multivariate curve resolution (MCR) to address potential coelution. The presented workflow is termed MF + MCR and was tested on pulsed elution-LC × LC-HRMS data from a wastewater effluent extract. The proposed workflow was benchmarked to the following three data processing strategies for mass spectra extraction: peak apex (PAM), using the MF approach alone, or using MCR without prior MF. The MF + MCR workflow identified 25 suspect compounds, compared to 23, 16, and 10 identified by MF, MCR, and PAM, respectively. The nine suspects that could not be identified using MCR compared to the MF + MCR all had low total signal contributions, i.e., low intensities compared to the TIC. This showed that adequate preprocessing prior to MCR is essential for trace level analysis. Additionally, it was shown that the MF + MCR workflow extracted statistically significantly purer mass spectra compared to PAM (p-value: 0.003) and MCR (p-value: 0.04) from a spiked blank sample. The results highlight that by utilizing the elution profiles in both chromatographic dimensions, clean mass spectra of analytes at trace levels measured in DIA can be extracted, allowing for more reliable compound identification compared to the workflows that were used for benchmarking.

DETERMINAÇÃO DE BIODIESEL METÍLICO DE DENDÊ EM MISTURAS COM DIESEL UTILIZANDO A RESOLUÇÃO MULTIVARIADA DE CURVAS COM MÍNIMOS QUADRADOS ALTERNADOS

DETERMINATION OF PALM METHYL BIODIESEL IN DIESEL BLENDS USING MULTIVARIATE CURVE RESOLUTION WITH ALTERNATING LEAST SQUARES. The increase in the average temperature of the Earth due to excess CO2 and other greenhouse gases has caused glaciers to melt and sea levels to rise. The burning of fossil fuels is one of the main causes of this problem, so the use of biofuels is important for mitigating this environmental impact. In Brazil, biodiesel is mixed with mineral; however, irregularities have been detected, reducing the expected positive effects. Therefore, this study proposes the use of multivariate curve resolution with alternating least squares (MCR-ALS) to quantify the content of methyl palm oil biodiesel in diesel blends and determine the concentration of the adulterant automotive lubricating oil in the blend. The model for quantifying biodiesel in diesel obtained R2 = 0.9997, root mean square error of calibration (RMSEC) = 0.56% (v v-1) and prediction (RMSEP) = 0.65% (v v-1), while the determination of adulteration in the biodiesel-diesel blend showed RMSEC = 0.67% (v v-1), RMSEP = 0.81% (v v-1) and R2 = 0.9992. These results demonstrate the ability of the models developed to predict both the content and quantity of possible adulterants in blends of palm oil biodiesel and mineral diesel, contributing to the quality control of this fuel.

Calibration free approaches for rapid polymorph discrimination via low frequency (THz) Raman spectroscopy.

Application of multivariate curve resolution to non-invasive Raman spectra has been investigated for rapid on-line analysis of crystallisation processes and high-throughput screening. Exploring quantification of mefenamic acid solid forms (form I, form II, and dimethylformamide solvate) from the Raman spectra indicated excellent agreement with off-line X-ray analysis.