Use Of Gaussian Mixture Models Research Articles

Efficient Storage and Importance Sampling for Fluorescent Reflectance

AbstractWe propose a technique for efficient storage and importance sampling of fluorescent spectral data. Fluorescence is fully described by a re‐radiation matrix, which for a given input wavelength indicates how much energy is re‐emitted at other wavelengths. However, such representation has a considerable memory footprint. To significantly reduce memory requirements, we propose the use of Gaussian mixture models for the representation of re‐radiation matrices. Instead of the full‐resolution matrix, we work with a set of Gaussian parameters that also allow direct importance sampling. Furthermore, if accuracy is of concern, a re‐radiation matrix can be used jointly with efficient importance sampling provided by the Gaussian mixture. In this paper, we present our pipeline for efficient storage of bispectral data and provide its extensive evaluation on a large set of bispectral measurements. We show that our method is robust and colour accurate even with its comparably minor memory requirements and that it can be seamlessly integrated into a standard Monte Carlo path tracer.

Measurement errors in Gaussian mixture models using high-dimensional air pollution constituents data

BACKGROUND AND AIM: An active area of current interest in air pollution epidemiology is characterizing the impact of air pollution constituents on health outcomes. Statistical analysis of air pollution data suffers from bias due to measurement error of these exposures, and this bias is exacerbated when the exposures are high-dimensional. It is well known that exposure to fine particles (PM2.5) can cause adverse health effects. However, the independent effects of PM2.5 constituents have not been as well investigated. METHODS: In this work, we develop a novel approach to correcting for the bias due to measurement errors associated with multiple correlated air pollutants with a complex correlated error structure through the use of Gaussian mixture models under the main / external validation design framework. The clusters are corrected for measurement error and then analyzed in a Cox model to investigate the impacts of different air pollutant profiles on time to even outcomes, such as all-cause mortality and cancer incidence. RESULTS: Extensive simulation studies are carried out to investigate the performance of the proposed method, and to explore sensitivity to departures from Gaussian assumptions. The method is illustrated in a study of air pollutants on all-cause mortality in the Nurses’ Health Study, with exposures validated in the RIOPA and MESA studies.

Reconstruction of Nuclear Ensemble Approach Electronic Spectra Using Probabilistic Machine Learning.

The theoretical prediction of molecular electronic spectra by means of quantum mechanical (QM) computations is fundamental to gain a deep insight into many photophysical and photochemical processes. A computational strategy that is attracting significant attention is the so-called Nuclear Ensemble Approach (NEA), that relies on generating a representative ensemble of nuclear geometries around the equilibrium structure and computing the vertical excitation energies (ΔE) and oscillator strengths (f) and phenomenologically broadening each transition with a line-shaped function with empirical full-width δ. Frequently, the choice of δ is carried out by visually finding the trade-off between artificial vibronic features (small δ) and over-smoothing of electronic signatures (large δ). Nevertheless, this approach is not satisfactory, as it relies on a subjective perception and may lead to spectral inaccuracies overall when the number of sampled configurations is limited due to an excessive computational burden (high-level QM methods, complex systems, solvent effects, etc.). In this work, we have developed and tested a new approach to reconstruct NEA spectra, dubbed GMM-NEA, based on the use of Gaussian Mixture Models (GMMs), a probabilistic machine learning algorithm, that circumvents the phenomenological broadening assumption and, in turn, the use of δ altogether. We show that GMM-NEA systematically outperforms other data-driven models to automatically select δ overall for small datasets. In addition, we report the use of an algorithm to detect anomalous QM computations (outliers) that can affect the overall shape and uncertainty of the NEA spectra. Finally, we apply GMM-NEA to predict the photolysis rate for HgBrOOH, a compound involved in Earth’s atmospheric chemistry.

Optimizing spatial distribution of watershed-scale hydrologic models using Gaussian Mixture Models

Common methods for spatial distribution, such as hydrologic response units, are subjective, time-consuming, and fail to capture the full range of basin attributes. Recent advances in statistical-learning techniques allow for new approaches to this problem. We propose the use of Gaussian Mixture Models (GMMs) for spatial distribution of hydrologic models. GMMs objectively select the set of modeling locations that best represent the distribution of watershed features relevant to the hydrologic cycle. We demonstrate this method in two hydrologically distinct headwater catchments of the Sierra Nevada and show that it meets or exceeds the performance of traditionally distributed models for multiple metrics across the water balance at a fraction of the time cost. Finally, we use univariate GMMs to identify the most-important drivers of hydrologic processes in a basin. The GMM method allows for more robust, objective, and repeatable models, which are critical for advancing hydrologic research and operational decision making.

The Use of Gaussian Mixture Models with Atmospheric Lagrangian Particle Dispersion Models for Density Estimation and Feature Identification

Atmospheric Lagrangian particle dispersion models, LPDM, simulate the dispersion of passive tracers in the atmosphere. At the most basic level, model output consists of the position of computational particles and the amount of mass they represent. In order to obtain concentration values, this information is then converted to a mass distribution via density estimation. To date, density estimation is performed with a nonparametric method so that output consists of gridded concentration data. Here we introduce the use of Gaussian mixture models, GMM, for density estimation. We compare to the histogram or bin counting method for a tracer experiment and simulation of a large volcanic ash cloud. We also demonstrate the use of the mixture model for automatic identification of features in a complex plume such as is produced by a large volcanic eruption. We conclude that use of a mixture model for density estimation and feature identification has potential to be very useful.

Enhancing the sensitivity of transient gravitational wave searches with Gaussian mixture models

Identifying the presence of a gravitational wave transient buried in non-stationary, non-Gaussian noise which can often contain spurious noise transients (glitches) is a very challenging task. For a given data set, transient gravitational wave searches produce a corresponding list of triggers that indicate the possible presence of a gravitational wave signal. These triggers are often the result of glitches mimicking gravitational wave signal characteristics. To distinguish glitches from genuine gravitational wave signals, search algorithms estimate a range of trigger attributes, with thresholds applied to these trigger properties to separate signal from noise. Here, we present the use of Gaussian mixture models, a supervised machine learning approach, as a means of modelling the multi-dimensional trigger attribute space. We demonstrate this approach by applying it to triggers from the coherent Waveburst search for generic bursts in LIGO O1 data. By building Gaussian mixture models for the signal and background noise attribute spaces, we show that we can significantly improve the sensitivity of the coherent Waveburst search and strongly suppress the impact of glitches and background noise, without the use of multiple search bins as employed by the original O1 search. We show that the detection probability is enhanced by a factor of 10, leading enhanced statistical significance for gravitational wave signals such as GW150914.

Augmenting machine learning photometric redshifts with Gaussian mixture models

ABSTRACT Wide-area imaging surveys are one of the key ways of advancing our understanding of cosmology, galaxy formation physics, and the large-scale structure of the Universe in the coming years. These surveys typically require calculating redshifts for huge numbers (hundreds of millions to billions) of galaxies – almost all of which must be derived from photometry rather than spectroscopy. In this paper, we investigate how using statistical models to understand the populations that make up the colour–magnitude distribution of galaxies can be combined with machine learning photometric redshift codes to improve redshift estimates. In particular, we combine the use of Gaussian mixture models with the high-performing machine-learning photo-z algorithm GPz and show that modelling and accounting for the different colour–magnitude distributions of training and test data separately can give improved redshift estimates, reduce the bias on estimates by up to a half, and speed up the run-time of the algorithm. These methods are illustrated using data from deep optical and near-infrared data in two separate deep fields, where training and test data of different colour–magnitude distributions are constructed from the galaxies with known spectroscopic redshifts, derived from several heterogeneous surveys.

Deep generative variational autoencoding for replay spoof detection in automatic speaker verification

Automatic speaker verification (ASV) systems are highly vulnerable to presentation attacks, also called spoofing attacks. Replay is among the simplest attacks to mount — yet difficult to detect reliably. The generalization failure of spoofing countermeasures (CMs) has driven the community to study various alternative deep learning CMs. The majority of them are supervised approaches that learn a human-spoof discriminator. In this paper, we advocate a different, deep generative approach that leverages from powerful unsupervised manifold learning in classification. The potential benefits include the possibility to sample new data, and to obtain insights to the latent features of genuine and spoofed speech. To this end, we propose to use variational autoencoders (VAEs) as an alternative backend for replay attack detection, via three alternative models that differ in their class-conditioning. The first one, similar to the use of Gaussian mixture models (GMMs) in spoof detection, is to train independently two VAEs — one for each class. The second one is to train a single conditional model (C-VAE) by injecting a one-hot class label vector to the encoder and decoder networks. Our final proposal integrates an auxiliary classifier to guide the learning of the latent space. Our experimental results using constant-Q cepstral coefficient (CQCC) features on the ASVspoof 2017 and 2019 physical access subtask datasets indicate that the C-VAE offers substantial improvement in comparison to training two separate VAEs for each class. On the 2019 dataset, the C-VAE outperforms the VAE and the baseline GMM by an absolute 9 - 10% in both equal error rate (EER) and tandem detection cost function (t-DCF) metrics. Finally, we propose VAE residuals — the absolute difference of the original input and the reconstruction as features for spoofing detection. The proposed frontend approach augmented with a convolutional neural network classifier demonstrated substantial improvement over the VAE backend use case.

Gaussian mixture models for blended photometric redshifts

ABSTRACTFuture cosmological galaxy surveys such as the Large Synoptic Survey Telescope (LSST) will photometrically observe very large numbers of galaxies. Without spectroscopy, the redshifts required for the analysis of these data will need to be inferred using photometric redshift techniques that are scalable to large sample sizes. The high number density of sources will also mean that around half are blended. We present a Bayesian photometric redshift method for blended sources that uses Gaussian mixture models to learn the joint flux–redshift distribution from a set of unblended training galaxies, and Bayesian model comparison to infer the number of galaxies comprising a blended source. The use of Gaussian mixture models renders both of these applications computationally efficient and therefore suitable for upcoming galaxy surveys.

An Automatic Method for the Segmentation and Classification of Imminent Labor Contraction From Electrohysterograms

Preterm birth is the first cause of perinatal morbidity and mortality. Despite continuous clinical routine improvements, the preterm rate remains steady. Moreover, the specificity of the early diagnosis stays poor as many hospitalized women for preterm delivery threat finally deliver at term. In this context, the use of electrohysterograms may increase the sensitivity and the specificity of early diagnosis of preterm labor. This paper proposes a clinical application of electrohysterogram processing for the classification of patients as prone to deliver within a week or later. The approach relies on non-linear correlation analysis for the contraction bursts extraction and uses computation of various features combined with the use of Gaussian mixture models for their classification. The method is tested on a new dataset of 68 records collected on women hospitalized for preterm delivery threat. This paper presents promising results for the automatic segmentation of the contraction and a classification sensitivity, specificity, and accuracy of, respectively, 80.7%, 76.3%, and 76.2%. These results are in accordance with the gold standards but have the advantage to be non-invasive and could be performed at home. Diagnosis of imminent labor is possible by electrohysterography recording and may help in avoiding over-medication and in providing better cares to at-risk pregnant women.

Joint Channel-Estimation and Equalization of Single-Carrier Systems via Bilinear AMP

We propose a novel soft-input soft-output equalizer for single-carrier transmissions over unknown frequency-selective block-fading channels. Our equalizer leverages the recently proposed parametric bilinear generalized approximate message passing algorithm for joint channel-estimation and symbol-detection, and exploits fast Fourier transform (FFT)-processing to achieve a per-symbol complexity that grows only logarithmically in the channel delay-spread. Furthermore, it supports the use of Gaussian mixture models to support the approximately sparse nature of wideband wireless channel responses. Numerical experiments, conducted using physically motivated Saleh–Valenzuela channel models, show that the proposed approach achieves channel normalized mean square error and bit error rate that are significant improved over existing turbo frequency-domain equalization approaches for unknown channels. Additional experiments show that the proposed scheme facilitates much higher spectral efficiencies than sparse deconvolution methods based on convex relaxation.

PREDICTION OF BUILDING FLOORPLANS USING LOGICAL AND STOCHASTIC REASONING BASED ON SPARSE OBSERVATIONS

Abstract. This paper introduces a novel method for the automatic derivation of building floorplans and indoor models. Our approach is based on a logical and stochastic reasoning using sparse observations such as building room areas. No further sensor observations like 3D point clouds are needed. Our method benefits from an extensive prior knowledge of functional dependencies and probability density functions of shape and location parameters of rooms depending on their functional use. The determination of posterior beliefs is performed using Bayesian Networks. Stochastic reasoning is complex since the problem is characterized by a mixture of discrete and continuous parameters that are in turn correlated by non-linear constraints. To cope with this kind of complexity, the proposed reasoner combines statistical methods with constraint propagation. It generates a limited number of hypotheses in a model-based top-down approach. It predicts floorplans based on a-priori localised windows. The use of Gaussian mixture models, constraint solvers and stochastic models helps to cope with the a-priori infinite space of the possible floorplan instantiations.

SPEAKER IDENTIFICATION BY AGGREGATING GAUSSIAN MIXTURE MODELS (GMMs) BASED ON UNCORRELATED MFCC-DERIVED FEATURES

For solving speaker identification problems, the approach proposed by Reynolds [IEEE Signal Process. Lett.2 (1995) 46–48], using Gaussian Mixture Models (GMMs) based on Mel Frequency Cepstral Coefficients (MFCCs) as features, is one of the most effective available in the literature. The use of GMMs for modeling speaker identity is motivated by the interpretation that the Gaussian components represent some general speaker-dependent spectral shapes, and also by the capability of Gaussian mixtures to model arbitrary densities. In this work, we have initially illustrated, with the help of a new bilingual speech corpus, how the well-known principal component transformation, in conjunction with the principle of classifier combination can be used to enhance the performance of the MFCC-GMM speaker recognition systems significantly. Subsequently, we have emphatically and rigorously established the same using the benchmark speech corpus NTIMIT. A significant outcome of this work is that the proposed approach has the potential to enhance the performance of any speaker recognition system based on correlated features.

Towards Robust 3D Face Verification Using Gaussian Mixture Models

This paper focuses on the use of Gaussian Mixture models (GMM) for 3D face verification. A special interest is taken in practical aspects of 3D face verification systems, where all steps of the verification procedure need to be automated and no meta-data, such as pre-annotated eye/nose/mouth positions, is available to the system. In such settings the performance of the verification system correlates heavily with the performance of the employed alignment (i.e., geometric normalization) procedure. We show that popular holistic as well as local recognition techniques, such as principal component analysis (PCA), or Scale-invariant feature transform (SIFT)-based methods considerably deteriorate in their performance when an “imperfect” geometric normalization procedure is used to align the 3D face scans and that in these situations GMMs should be preferred. Moreover, several possibilities to improve the performance and robustness of the classical GMM framework are presented and evaluated: i) explicit inclusion of spatial information, during the GMM construction procedure, ii) implicit inclusion of spatial information during the GMM construction procedure and iii) on-line evaluation and possible rejection of local feature vectors based on their likelihood. We successfully demonstrate the feasibility of the proposed modifications on the Face Recognition Grand Challenge data set.

Probabilistic load flow with non-Gaussian correlated random variables using Gaussian mixture models

This study proposes the use of Gaussian mixture models to represent non-Gaussian correlated input variables, such as wind power output or aggregated load demands in the probabilistic load flow problem. The algorithm calculates the marginal distribution of any bus voltage or power flow as a sum of Gaussian components obtained from multiple weighted least square runs. The number of trials depends on the number of Gaussian components used to model each input random variable. Monte Carlo simulations are used to compare the approximations. The effect of correlation between variables is taken into consideration in both formulations. The main advantage of the Gaussian components method is that the probability density functions of any variable is directly obtained. Test results in the 14-bus system and the 57-bus system provide a broad explanation of the advantages and constraints of the approximations, particularly in presence of correlated variables.

SiMOR: Single Moving Object Recognition

Automatic moving object detection and tracking is very important task in video surveillance applications. In the present work the well known background subtraction model and use of Gaussian Mixture Models (GMM) have been used to implement a robust automated single object tracking system. In this implementation, background subtraction on subtracting consecutive frame-by-frame basis for moving object detection is done. Once the object has been detected it is tracked by employing an efficient GMM technique. After successful completion of tracking, moving object recognition of those objects using well known Principal Component Analysis (PCA), which is used for extracting features and Manhattan based distance metric is used for subsequent classification purpose. The system is capable of handling entry and exit of an object. Such a tracking system is cost effective and can be used as an automated video conferencing system and also has applications like human tracking, vehicles monitoring, and event recognition for video surveillance. The proposed algorithm was tested on standard database on complex environments and the results were satisfactory.

Automatic classification of vocalizations with Gaussian Mixture Models and Hidden Markov Models

The automatic classification of marine mammal sounds is very attractive as a means of assessing massive quantities of recorded data, freeing humans and offering rigorous and consistent output. Calculations on a set of vocalizations of Northern Resident killer whales using Dynamic Time Warping have been reported recently. [Brown, J. C., and P.J.O. Miller, ''Automatic classification of killer whale vocalizations using dynamic time warping,'' J. Acoust. Soc. Am. 122, 1201‐1207, (2007).] Since this method requires the time‐consuming pre‐processing measurement of the frequency contours, we have explored the use of Gaussian Mixture Models (GMM) and Hidden Markov Models (HMM). These methods can be applied directly to time‐frequency decompositions of the recorded signals. Calculations have been made on a set of 75 calls previously classified perceptually into 7 call types. Preliminary results give an agreement of roughly 85% agreement with the perceptual classification for the GMM and over 90% for an HMM.

The question of disguised voice

State of the art automatic speaker recognition systems show very good results in the discrimination between different speakers under controlled recording conditions. In a forensic context, the conditions are uncontrolled and voice can be disguised. In cases of terrorism claim, extortion or kidnapping, it is of great interest for offenders to conceal their identity. Voice disguise is an important constraint to speaker discrimination. Some disguises produce a great variation of parameters and change the perception of an identity. The main risk is to confound a disguised voice and a normal voice and accuse an innocent individual. This paper proposes on one hand to present the impact of voice disguise on automatic speaker recognition and, on the other hand a statistical study in order to detect and identify four disguises among the most common. The first step consists in extracting features and the second step to classify them. MFCC (Mel Frequency Cepstral Coefficient) are considered as features and different classification algorithms have been tested. The studied disguises are based on a deliberated and non electronic way. The proposed analysis of disguised voice classification provides interesting results in detection by the use of SVM (Support Vector Machine) and in identification by the use of GMM (Gaussian mixture models).

A Gaussian Mixture Model Based Classification Scheme for Myoelectric Control of Powered Upper Limb Prostheses

This paper introduces and evaluates the use of Gaussian mixture models (GMMs) for multiple limb motion classification using continuous myoelectric signals. The focus of this work is to optimize the configuration of this classification scheme. To that end, a complete experimental evaluation of this system is conducted on a 12 subject database. The experiments examine the GMMs algorithmic issues including the model order selection and variance limiting, the segmentation of the data, and various feature sets including time-domain features and autoregressive features. The benefits of postprocessing the results using a majority vote rule are demonstrated. The performance of the GMM is compared to three commonly used classifiers: a linear discriminant analysis, a linear perceptron network, and a multilayer perceptron neural network. The GMM-based limb motion classification system demonstrates exceptional classification accuracy and results in a robust method of motion classification with low computational load.

Robust text-independent speaker identification using Gaussian mixture speaker models

This paper introduces and motivates the use of Gaussian mixture models (GMM) for robust text-independent speaker identification. The individual Gaussian components of a GMM are shown to represent some general speaker-dependent spectral shapes that are effective for modeling speaker identity. The focus of this work is on applications which require high identification rates using short utterance from unconstrained conversational speech and robustness to degradations produced by transmission over a telephone channel. A complete experimental evaluation of the Gaussian mixture speaker model is conducted on a 49 speaker, conversational telephone speech database. The experiments examine algorithmic issues (initialization, variance limiting, model order selection), spectral variability robustness techniques, large population performance, and comparisons to other speaker modeling techniques (uni-modal Gaussian, VQ codebook, tied Gaussian mixture, and radial basis functions). The Gaussian mixture speaker model attains 96.8% identification accuracy using 5 second clean speech utterances and 80.8% accuracy using 15 second telephone speech utterances with a 49 speaker population and is shown to outperform the other speaker modeling techniques on an identical 16 speaker telephone speech task.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>