Source Separation Approach Research Articles

Unveiling Hidden Sources of Dynamic Functional Connectome through a Novel Regularized Blind Source Separation Approach

Abstract The investigation of the brain’s functional connectome and its dynamic changes can provide valuable insights into brain organization and its reconfiguration. However, the analysis of dynamic functional connectivity (dFC) using functional magnetic resonance imaging (fMRI) faces major challenges, including the high dimensionality of brain networks, unknown latent sources underlying observed dFC, and the large number of brain connections that increase the risk of spurious findings. In this paper, we propose a new regularized blind source separation (BSS) method called dyna-LOCUS to address these challenges. dyna-LOCUS decomposes observed dFC measures to reveal latent source connectivity traits and their dynamic temporal expression profiles. By utilizing low-rank factorization and novel regularizations, dyna-LOCUS achieves efficient and reliable mapping of connectivity traits underlying the dynamic brain functional connectome, characterizes temporal changes of the connectivity traits that contribute to the reconfiguration in the observed dFC, and generates parsimonious and interpretable results in identifying whole-brain dFC states. We introduce a highly efficient iterative Node-Rotation algorithm that solves the non-convex optimization problem for learning dyna-LOCUS. Simulation studies demonstrate the advantages of our proposed method. Application of dyna-LOCUS to the Philadelphia Neurodevelopmental Cohort (PNC) study unveils latent connectivity traits and key brain connections and regions driving each of these neural circuits, reveals temporal expressions and interactions of these connectivity traits, and generates new findings regarding gender differences in the neurodevelopment of an executive function-related connectivity trait.

An improved blind Gaussian source separation approach based on generalized Jaccard similarity

An improved blind Gaussian source separation approach based on generalized Jaccard similarity

Adaptive DBSCAN Clustering and GASA Optimization for Underdetermined Mixing Matrix Estimation in Fault Diagnosis of Reciprocating Compressors.

Underdetermined blind source separation (UBSS) has garnered significant attention in recent years due to its ability to separate source signals without prior knowledge, even when sensors are limited. To accurately estimate the mixed matrix, various clustering algorithms are typically employed to enhance the sparsity of the mixed matrix. Traditional clustering methods require prior knowledge of the number of direct signal sources, while modern artificial intelligence optimization algorithms are sensitive to outliers, which can affect accuracy. To address these challenges, we propose a novel approach called the Genetic Simulated Annealing Optimization (GASA) method with Adaptive Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering as initialization, named the CYYM method. This approach incorporates two key components: an Adaptive DBSCAN to discard noise points and identify the number of source signals and GASA optimization for automatic cluster center determination. GASA combines the global spatial search capabilities of a genetic algorithm (GA) with the local search abilities of a simulated annealing algorithm (SA). Signal simulations and experimental analysis of compressor fault signals demonstrate that the CYYM method can accurately calculate the mixing matrix, facilitating successful source signal recovery. Subsequently, we analyze the recovered signals using the Refined Composite Multiscale Fuzzy Entropy (RCMFE), which, in turn, enables effective compressor connecting rod fault diagnosis. This research provides a promising approach for underdetermined source separation and offers practical applications in fault diagnosis and other fields.

Unsupervised neural decoding of signals recorded by thin-film electrode arrays implanted in muscles using autoencoding with a physiologically derived optimisation criterion

We present an autoencoder that learns the discharge times of individual motor units (MU) by processing multichannel electromyogram (EMG) recordings obtained by electrode arrays implanted in muscles, thereby providing a neural interface with the spinal cord. To this end, after preprocessing of the EMG via convolutive sphering, the encoder is constrained to apply an orthogonal transformation to the spatial data and the optimisation criterion enforces a temporal sparsity constraint. These constraints are based on the theoretical modelling of EMG generation. This decomposition method was evaluated on both simulated and experimental signals. For simulated signals, with a ratio of 1.5 between observations and sources, the average detection accuracy was 94% for up to 60 sources for SNR 20 dB. Moreover, the detection accuracy did not significantly decrease when decreasing the SNR to values as low as 0 dB. Experimental signals were collected in humans with thin-film invasive electrodes implanted in the tibialis anterior muscle. The results showed an accuracy on experimental data >90%. Moreover, the proposed method outperformed a state-of-the-art blind source separation approach in terms of the number of reliably detected motor units. In conclusion, we have proposed the first fully unsupervised neural network approach to the problem of neural decoding of intramuscular EMG time series by translating the available theoretical knowledge on the signal properties into an autoencoding architecture tailored to the decomposition problem.

Direction specific ambisonics source separation with end-to-end deep learning

Ambisonics is a scene-based spatial audio format that has several useful features compared to object-based formats, such as efficient whole scene rotation and versatility. However, it does not provide direct access to the individual source signals, so that these have to be separated from the mixture when required. Typically, this is done with linear spherical harmonics (SH) beamforming. In this paper, we explore deep-learning-based source separation on static Ambisonics mixtures. In contrast to most source separation approaches, which separate a fixed number of sources of specific sound types, we focus on separating arbitrary sound from specific directions. Specifically, we propose three operating modes that combine a source separation neural network with SH beamforming: refinement, implicit, and mixed mode. We show that a neural network can implicitly associate conditioning directions with the spatial information contained in the Ambisonics scene to extract specific sources. We evaluate the performance of the three proposed approaches and compare them to SH beamforming on musical mixtures generated with the musdb18 dataset, as well as with mixtures generated with the FUSS dataset for universal source separation, under both anechoic and room conditions. Results show that the proposed approaches offer improved separation performance and spatial selectivity compared to conventional SH beamforming.

Joint Graph Learning and Blind Separation of Smooth Graph Signals Using Minimization of Mutual Information and Laplacian Quadratic Forms

The smoothness of graph signals has found desirable real applications for processing irregular (graph-based) signals. When the latent sources of the mixtures provided to us as observations are smooth graph signals, it is more efficient to use graph signal smoothness terms along with the classic independence criteria in Blind Source Separation (BSS) approaches. In the case of underlying graphs being known, Graph Signal Processing (GSP) provides valuable tools; however, in many real applications, these graphs can not be well-defined <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a priori</i> and need to be learned from data. In this paper, a GSP-based approach for joint Graph Learning (GL) and BSS of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">smooth</i> graph signal sources is proposed, which, to the best of our knowledge, has not been addressed. The minimization of Mutual Information (MI) is exploited in our work because of its desirable advantages, such as having no local minimum and exact (not approximative) modeling of statistical independence. As well as investigating the convergence analysis and identifiability conditions and deriving Cramér-Rao Lower Bound (CRLB), comprehensive experiments of generating (cyclostationary) graph signal mixtures are provided to show the superiority over classic and also GSP-based BSS approaches in the case of the latent sources being <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">smooth</i> on <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">unknown</i> underlying graphs. Based on the provided experimental results, our proposed method performs quite robustly in even high noise levels in both terms of source separation and recovery of underlying graphs.

A Novel Blind Image Source Separation Using Hybrid Firefly Particle Swarm Optimization Algorithm

Signal and image separation are extensively used in numerous imaging applications and communication systems. In this paper, a novel Blind Source Separation (BSS) approach, based on the Hybrid Firefly Particle Swarm Optimization (HFPSO), is proposed for separating mixed images. This approach processes the observed source without any prior knowledge about the model and the statistics of the source signal. The proposed method presents high robustness against local minima and converges quickly to the global minimum. Via numerical simulations, the proposed approach is tested and validated in comparison with standard Particle Swarm Optimization (PSO), Robust Independent Component Analysis (RobustICA), and Artificial Bee Colony (ABC) algorithms. The obtained results show that the presented technique outperforms the existing ones in terms of quality of image separation, the Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM). Moreover, the obtained results demonstrate that our approach provides also promising results in image separation from noisy mixtures.

Bayesian Spatial Blind Source Separation via the Thresholded Gaussian Process

Blind source separation (BSS) aims to separate latent source signals from their mixtures. For spatially dependent signals in high-dimensional and large-scale data, such as neuroimaging, most existing BSS methods do not take into account the spatial dependence and the sparsity of the latent source signals. To address these major limitations, we propose a Bayesian spatial blind source separation (BSP-BSS) approach for neuroimaging data analysis. We assume the expectation of the observed images as a linear mixture of multiple sparse and piece-wise smooth latent source signals, for which we construct a new class of Bayesian nonparametric prior models by thresholding Gaussian processes. We assign the vMF priors to mixing coefficients in the model. Under some regularity conditions, we show that the proposed method has several desirable theoretical properties including the large support for the priors, the consistency of joint posterior distribution of the latent source intensity functions and the mixing coefficients, and the selection consistency on the number of latent sources. We use extensive simulation studies and an analysis of the resting-state fMRI data in the Autism Brain Imaging Data Exchange (ABIDE) study to demonstrate that BSP-BSS outperforms the existing method for separating latent brain networks and detecting activated brain activation in the latent sources. Supplementary materials for this article are available online.

Fetal ECG Signal Extraction From Long-Term Abdominal Recordings Based on Adaptive QRS Removal and Joint Blind Source Separation

Despite wearable electronic devices having enormous potential applications, the recording quality has not sufficiently met the requirements for fetal disease detection. In this article, we aim to propose a novel framework for fetal electrocardiography (ECG) extraction, where the hybrid approach mainly considers the combination of a novel adaptive maternal QRS removal (AQRSR) algorithm and a tensor-based joint blind source separation (JBSS) approach. To adapt to the dynamic change of the ECG signal and suppress the powerful maternal heartbeat, AQRSR is explored with an adaptive template, which is generated by a model consisting of a generator, discriminator, and transformer. By replacing the oldest segmentation with the newest cycle, the approach is capable of quickly adapting to the estimated output to match the new input. In addition, due to the significant crossover between the target signal and the noise, even if the maternal ECG removal processing is implemented, the residual signal inevitably contains fetal ECG, part of maternal ECG, and noise. As a consequence of this, a novel JBSS that incorporates tensor decomposition is formulated to separate fetal heartbeat from long-term recordings contaminated with maternal movements noise and heartbeat interference. By dividing the collected data into multiple segments, this method can extract fetal ECG signals by exploiting the signals that are not only statistically independent within each segment but can be dependent on different segments. Experimental results show that even if the recorded signals are contaminated by the noise, the proposed framework has comparable better performance on fetal ECG extraction.

X-ray fluorescence scanning of soft and wet-sediment cores in terrestrial environments; A robust blind source separation approach

X-ray fluorescence (XRF) scanning of split cores using automated cores scanners is rapid but expensive and semi-quantitative method in sedimentary geochemistry. While searching for a low-cost and quantitative solution in environmental geochemistry, we investigated the application of hand-held XRF device in scanning on two siliciclastic-sediment cores from dam-reservoirs in Slovakia. The core-scan data were compared with XRF data from dry, powdered sample aliquots, grain size, water content, and magnetic susceptibility data, which were used as independent variables. The aim of the manuscript is to investigate the applicability of time-series statistical methods to reveal the physical and chemical signals from hand-held XRF scanning of wet-sediment cores. The geochemical data were processed using the centred log-ratio (clr) technique, and then by the Second-Order Blind Identification (SOBI) method, which is a subtype of robust Blind Source Separation (BSS) techniques commonly used to enhance the signal-to-noise ratio in time series analysis of data in medicine or acoustics. So far, BSS has been rarely applied in geochemistry. Raw element concentrations (ppm/wt%) of core scans and sediment powders are essentially incomparable, but their clr coefficients show statistically significant correlation for Al, P, K, Ca, Mn, Fe, Cu, As, Sr, Zr, and Pb, which reduces the difference between the two measurement methods and justifies previous attempts to calibrate the core-scanning data using log-ratio techniques. Scores of three of eight latent components (IC2, IC4, and IC7) found by SOBI in the core scan- and powder data show significant statistical correlation demonstrating the applicability of the core-scan technique. Their geochemical interpretation is based on the IC loadings by element clr, and the stratigraphic correlation with the sediment grain size and water content. These ICs explain geochemical variability due to grain-size contrast between sandy and silty layers, input of biogenic productivity-sensitive elements (Ca and P), eutrophication, and accumulation of anthropogenic elements. The robust BSS approach represents a promising method of processing of large datasets in environmental geochemistry.

Semi-Blind Source Separation with Learned Constraints

Blind source separation (BSS) algorithms are unsupervised methods, which are the cornerstone of hyperspectral data analysis by allowing for physically meaningful data decompositions. BSS problems being ill-posed, the resolution requires efficient regularization schemes to better distinguish between the sources and yield interpretable solutions. For that purpose, we investigate a semi-supervised source separation approach in which we combine a projected alternating least-square algorithm with a learning-based regularization scheme. In this article, we focus on constraining the mixing matrix to belong to a learned manifold by making use of generative models. Altogether, we show that this allows for an innovative BSS algorithm, with improved accuracy, which provides physically interpretable solutions. The proposed method, coined sGMCA, is tested on realistic hyperspectral astrophysical data in challenging scenarios involving strong noise, highly correlated spectra and unbalanced sources. The results highlight the significant benefit of the learned prior to reduce the leakages between the sources, which allows an overall better disentanglement.

A novel Bayesian blind source separation approach for extracting non-stationary and discontinuous components from structural health monitoring data

We propose a new method to explore the blind source separation (BSS) of heterogeneous structural health monitoring (SHM) data containing non-stationary and temporally discontinuous components in the framework of Bayesian inference. Specifically, Gaussian process (GP) with a specially defined time-varying kernel function is introduced to encode the prior information about the unknown sources. The time-varying kernel function encompasses a state indicator hyperparameter which enables the expressivity of intermittently active and inactive source signals and incorporates smooth switch and composite kernels catering for the interpretation of complex sources. With the likelihood function elicited from monitoring data alongside with the priors for sources, mixing matrix and noise, the unknown sources are extracted from the posterior distributions formalized by Bayes’ theorem, where a sequential Metropolis − Hasting sampling algorithm is adopted to numerically compute the source statistics in a high-dimensional realm. Numerical simulations demonstrate that the proposed method performs satisfactorily in (i) extracting intermittently active and inactive (abruptly appearing and disappearing) non-stationary source signals, (ii) estimating inconsistent levels of noise contaminated in different sensors, and (iii) handling unknown number of sources. In the verification using real-world strain monitoring data collected from the Tsing Ma Bridge carrying both highway and railway traffic, it is shown that the proposed method well extracts the railway-induced non-stationary strain component with intermittence, and the structural condition index formulated by the Bayesian dynamic linear model (BDLM) is more robust when adopting the separated highway-induced strain data than using the combined railway- and highway-induced strain data. The results obtained by the proposed method are also compared with those by the two most common BSS techniques — the independent component analysis (ICA) and second-order statistics (SOS) methods.

Hidden patterns of gene expression provide prognostic insight for colorectal cancer.

Cancer tissue samples contain cancer cells and non-cancer cells with each biopsied site containing distinct proportions of these populations. Consequently, assigning useful tumor subtypes based on gene expression measurements from clinical samples is challenging. We applied a blind source separation approach to extract cancer cell-intrinsic gene expression patterns within clinical tumor samples of colorectal cancer. After a blind source separation, we found that a cancer cell-intrinsic gene expression program unique to each patient exists in the "residual" expression profile remaining after separation of the gene expression data. We performed a consensus clustering analysis of the extracted gene expression profiles to identify novel and robust cancer cell-intrinsic subtypes. We validated the identified subtypes using an independent clinical gene expression dataset. The cancer cell-intrinsic subtypes are independent of biopsy site and provided prognostic information in addition to currently available clinical and molecular variables. After validating this approach in colorectal cancer, we further identified novel tumor subtypes with unique clinical information across multiple types of cancer. These cancer cell-intrinsic molecular subtypes provide novel prognostic value for clinical assessment of cancer.

Input motion extraction and closely-spaced modes recognition through partial system identification: An orthogonality-aided blind source separation approach

This research proposes an output-only procedure to extract seismic input motion and identify systems with severe closely-spaced modes incorporating a partial system identification approach. The proposed framework employs new methods to detect free vibration portion of the response histories and clustering the candidate points to find the first mode shape in a blind modal identification (BMI) procedure. The first mode frequency and damping along with mass matrix distribution of the system are calculated then using an orthogonality-based optimization process. The mentioned identified features are utilized next for the input motion extraction without any need to know the higher mode characteristics of the system. Next, higher mode characteristics are detected in succeeding steps considering the removal of each identified mode share from the total response at the end of each step. The perfect performance of the framework was proved against two simulated synthetic structures with mild and severe closely-spaced modes.

Noncontact Multi-Target Respiration Sensing Using SIMO Radar With UBSS Method

Effectively extracting vital signs from multiple targets has been challenging for non-contact vital signs monitoring. This letter presents a nonlinear projection-affinity propagation (NP-AP) clustering-based underdetermined blind source separation (UBSS) approach that can isolate respiratory signals of multiple targets in an underdetermined condition without predetermining the number of targets. The proposed algorithm can provide an accurate estimation of the mixing matrix and is robust to the noises. The experimental result with three human subjects demonstrates the effectiveness of the proposed method in a real scenario.

Joint Energy Disaggregation of Behind-the-Meter PV and Battery Storage: A Contextually Supervised Source Separation Approach

An increasing number of residential customers have installed hybrid rooftop solar battery systems (HRSBSs). Currently, most HRSBSs are installed behind-the-meter (BTM), where only customers’ net load is measured by smart meters. This invisibility poses significant challenges to the system operation. Disaggregating BTM PV generation and battery charging/discharging profile of customers can enhance the grid-edge observability. This article proposes an energy disaggregation method to jointly separate the PV generation and battery charging/discharging power from the net load. First, a Home Smart Battery Management model is built to simulate the battery charging/discharging profile. Second, an optimal disaggregation model is established based on a contextually supervised source separation method. Furthermore, the feature vectors of PV, load, and battery are extracted as the input of the disaggregation model. Case studies on two datasets from Australia and China show that the proposed method has a promising disaggregation performance.

A Unified Approach for Simultaneous Graph Learning and Blind Separation of Graph Signal Sources

Blind separation of the sources supported by graphs, i.e., graph signals, is a nascent and challenging Blind Source Separation (BSS) problem. In these cases, along with the statistical independence of the sources, additional dependency information can be interpreted from their graph structure, which provides exploiting different Graph Signal Processing (GSP) tools to improve the separation performance. To the best of our knowledge, in these cases, only GraDe and GraphJADE methods have been proposed to exploit the graph dependencies and/or GSP techniques to improve the separation quality rather than the conventional BSS methods. Despite the significant advantages of these graph-based methods, assuming that the underlying graphs are known is a serious drawback, especially in many real-world applications. To address this issue, in this paper, we propose a Unified objective function for GraphJADE with Graph Learning, namely U-GraphJADE-GL, and use the Block Coordinate Descent (BCD) to optimize it. In our proposed method, along with the separation task, the underlying graphs are learned simultaneously in an efficient manner. We compare the performance of the U-GraphJADE-GL with that of the GraDe with Graph Learning (U-GraDe-GL) and the conventional BSS methods in the BSS task and also analyze the GL performance. Besides, as well as the theoretical and experimental convergence analysis, we derive/state the Cram&#x00E9;r-Rao bound (CRB) on the estimation of the mixing and unmixing matrices and also on the attainable Interference-to-Source Ratio (ISR) and compare the asymptotic performance of the proposed method with that of the optimal CRB estimators. To investigate the applicability in real applications, the performance of the proposed method is also compared with that of the abovementioned algorithms for denoising the epileptic Electroencephalogram (EEG) signals and also for the audio speech source separation task. The results show the superiority of the proposed algorithm compared to the other methods in the case of BSS of graph signals with unknown graphs.

Convolutive Transfer Function-Based Multichannel Nonnegative Matrix Factorization for Overdetermined Blind Source Separation

Most multichannel blind source separation (BSS) approaches rely on a spatial model to encode the transfer functions from sources to microphones and a source model to encode the source power spectral density. The rank-1 spatial model has been widely exploited in independent component analysis (ICA), independent vector analysis (IVA), and independent low-rank matrix analysis (ILRMA). The full-rank spatial model is also considered in many BSS approaches, such as full-rank spatial covariance matrix analysis (FCA), multichannel nonnegative matrix factorization (MNMF), and FastMNMF, which can improve the separation performance in the case of long reverberation times. This paper proposes a new MNMF framework based on the convolutive transfer function (CTF) for overdetermined BSS. The time-domain convolutive mixture model is approximated by a frequency-wise convolutive mixture model instead of the widely adopted frequency-wise instantaneous mixture model. The iterative projection algorithm is adopted to estimate the demixing matrix, and the multiplicative update rule is employed to estimate nonnegative matrix factorization (NMF) parameters. Finally, the source image is reconstructed using a multichannel Wiener filter. The advantages of the proposed method are twofold. First, the CTF approximation enables us to use a short window to represent long impulse responses. Second, the full-rank spatial model can be derived based on the CTF approximation and slowly time-variant source variances, and close relationships between the proposed method and ILRMA, FCA, MNMF and FastMNMF are revealed. Extensive experiments show that the proposed algorithm achieves a higher separation performance than ILRMA and FastMNMF in reverberant environments.

Blind Source Separation for MT-InSAR Analysis With Structural Health Monitoring Applications

Monitoring large areas of the Earth&#x0027;s surface is possible thanks to the availability of remote sensing data obtained by a large collection of diverse satellites orbiting the Earth. Specifically, multitemporal interferometric synthetic aperture radar (MT-InSAR) techniques supply the structural community with time series of line-of-sight displacements of terrain or structures, such as bridges or buildings, resulting from causes such as thermal expansion, contraction, and terrain deformation. The analysis of the different deformation signals observed is crucial in order to identify the different phenomena that cause the deformation. In this article, we explore the possibility of applying blind source separation algorithms to MT-InSAR, with the aim of developing methods towards automatic identification of different deformation patterns both in buildings and structures. We validate the proposed methodology using both synthetically generated datasets and real MT-InSAR data. We also provide a comparison with other similar methods. Our results demonstrate that InSAR time-series analysis can benefit from the use of the proposed blind source separation approach. Furthermore, the proposed technique is robust to the noise which is usually present in MT-InSAR data. This opens the door for monitoring infrastructure at scale over very large areas, helping to monitor civil infrastructures, and providing relevant insights to asset owners regarding the performance of such structures over time.

A multi-step blind source separation approach for the attenuation of artifacts in mobile high-density electroencephalography data

Objective. Electroencephalography (EEG) is a widely used technique to address research questions about brain functioning, from controlled laboratorial conditions to naturalistic environments. However, EEG data are affected by biological (e.g. ocular, myogenic) and non-biological (e.g. movement-related) artifacts, which—depending on their extent—may limit the interpretability of the study results. Blind source separation (BSS) approaches have demonstrated to be particularly promising for the attenuation of artifacts in high-density EEG (hdEEG) data. Previous EEG artifact removal studies suggested that it may not be optimal to use the same BSS method for different kinds of artifacts. Approach. In this study, we developed a novel multi-step BSS approach to optimize the attenuation of ocular, movement-related and myogenic artifacts from hdEEG data. For validation purposes, we used hdEEG data collected in a group of healthy participants in standing, slow-walking and fast-walking conditions. During part of the experiment, a series of tone bursts were used to evoke auditory responses. We quantified event-related potentials (ERPs) using hdEEG signals collected during an auditory stimulation, as well as the event-related desynchronization (ERD) by contrasting hdEEG signals collected in walking and standing conditions, without auditory stimulation. We compared the results obtained in terms of auditory ERP and motor-related ERD using the proposed multi-step BSS approach, with respect to two classically used single-step BSS approaches. Main results. The use of our approach yielded the lowest residual noise in the hdEEG data, and permitted to retrieve stronger and more reliable modulations of neural activity than alternative solutions. Overall, our study confirmed that the performance of BSS-based artifact removal can be improved by using specific BSS methods and parameters for different kinds of artifacts. Significance. Our technological solution supports a wider use of hdEEG-based source imaging in movement and rehabilitation studies, and contributes to the further development of mobile brain/body imaging applications.