Gaussian Mixture Model Representation Research Articles

Image Synthesis for Solar Flare Prediction

Solar flare prediction is a topic of interest to many researchers owing to the potential of solar flares to affect various technological systems, both terrestrial and in orbit. In recent years, the forecasting task has become progressively more reliant on data-driven computations and machine-learning algorithms. Although these efforts have improved solar flare predictions, they still falter in doing so for large solar flares, in particular under operational conditions, since large-flare data are very scarce and labeled data are heavily imbalanced. In this work, we seek to address this fundamental issue and present a scheme for generating synthetic magnetograms to reduce the imbalance in the data. Our method consists of (1) synthetic oversampling of line-of-sight magnetograms using Gaussian mixture model representation, followed by (2) a global optimization technique to ensure consistency of both physical features and flare precursors, and (3) the mapping of the generated representations to realistic magnetogram images using deep generative models. We show that these synthetically generated data indeed improve the capacity of solar flare prediction models and that, when tested on such a state-of-the-art model, it significantly enhances its forecasting performance, achieving an F1-score as high as 0.43 ± 0.08 and a true skill statistic of 0.64 ± 0.10 for X-class flares in the 24 hr operational solar flare data split.

Structured Gaussian Process Regression of Music Mood

Modeling the music mood has wide applications in music categorization, retrieval, and recommendation systems; however, it is challenging to computationally model the affective content of music due to its subjective nature. In this work, a structured regression framework is proposed to model the valence and arousal mood dimensions of music using a single regression model at a linear computational cost. To tackle the subjectivity phenomena, a confidence-interval based estimated consensus is computed by modeling the behavior of various annotators (e.g. biased, adversarial) and is shown to perform better than using the average annotation values. For a compact feature representation of music clips, variational Bayesian inference is used to learn the Gaussian mixture model representation of acoustic features and chord-related features are used to improve the valence estimation by probing the chord progressions between chroma frames. The dimensionality of features is further reduced using an adaptive version of kernel PCA. Using an efficient implementation of twin Gaussian process for structured regression, the proposed work achieves a significant improvement in R2 for arousal and valence dimensions relative to state-of-the-art techniques on two benchmark datasets for music mood estimation.

Modeling acoustic cues to distinctive features in a lexical speech analysis system

Acoustic cues are robust elements that can be used to infer information contained in the speech signal, such as underlying linguistic distinctive features and the words intended by the speaker (Stevens JASA 2002). Yet, most current automatic speech recognition systems do not take advantage of a feature-cue-based framework for signal analysis. In this project, a set of common acoustic cues has been explicitly modeled by Gaussian mixture models. This set of acoustic cues can provide evidence for the overall phoneme and word sequences of an utterance. The extracted cues and their values can also determine a speaker’s linguistic production pattern, i.e., the systematic context-governed modifications in surface-phonetic form that occur pervasively in conversational speech. The simple Gaussian mixture model representation structure reduces the need for extensive amounts of training data, in contrast to conventional schemes based on large neural networks.

A Gaussian Mixture Model Representation of Endmember Variability in Hyperspectral Unmixing

Hyperspectral unmixing while considering endmember variability is usually performed by the normal compositional model, where the endmembers for each pixel are assumed to be sampled from unimodal Gaussian distributions. However, in real applications, the distribution of a material is often not Gaussian. In this paper, we use Gaussian mixture models (GMM) to represent endmember variability. We show, given the GMM starting premise, that the distribution of the mixed pixel (under the linear mixing model) is also a GMM (and this is shown from two perspectives). The first perspective originates from random variable transformations and gives a conditional density function of the pixels given the abundances and GMM parameters. With proper smoothness and sparsity prior constraints on the abundances, the conditional density function leads to a standard maximum a posteriori (MAP ) problem which can be solved using generalized expectation maximization. The second perspective originates from marginalizing over the endmembers in the GMM, which provides us with a foundation to solve for the endmembers at each pixel. Hence, compared to the other distribution based methods, our model can not only estimate the abundances and distribution parameters, but also the distinct endmember set for each pixel. We tested the proposed GMM on several synthetic and real datasets, and showed its potential by comparing it to current popular methods.

Trajectory learning and reproduction for differential drive mobile robots based on GMM/HMM and dynamic time warping using learning from demonstration framework

In this paper, we present new Learning from Demonstration-based algorithm that generalizes and extracts relevant features of desired motion trajectories for differential drive mobile robots. The algorithm is tested through series of simulations and real world experiments in which desired task is demonstrated by the human teacher while teleoperating the mobile robot in the working environment. In the first step of the developed method, Gaussian Mixture Model (GMM) is built for incremental motions of the mobile robot between two consecutive poses. After this, the hidden Markov model is used to capture transitions between states (temporal variations of the data between clusters) which are missing from static GMM representation. Generalization of the motion is achieved by using the concept of keyframes, defined as points in which significant changes between GMM/HMM states occur. In the second step, the resulting GMM/HMM representation is used to generate optimal state sequences for each demonstration and to temporally align them, using 1D dynamic time warping, with respect to the one most consistent with the GMM/HMM model. This phase implies extraction of keyframes along all state sequences and projecting them into control space, in which controls are aligned in time as well. Finally, the generalized controls are obtained by averaging over all controls at the keyframes; simple piecewise cubic spline method is used for interpolation between generated control values. The main advantage of the developed algorithm is its ability to learn and generalize from all demonstrated examples which results in high quality reproductions of the motion. The proposed approach is verified both in simulated environment and using real mobile robot.

Life-Threatening Arrhythmia Verification in ICU Patients Using the Joint Cardiovascular Dynamical Model and a Bayesian Filter

In this paper, a novel nonlinear joint dynamical model is presented, which is based on a set of coupled ordinary differential equations of motion and a Gaussian mixture model representation of pulsatile cardiovascular (CV) signals. In the proposed framework, the joint interdependences of CV signals are incorporated by assuming a unique angular frequency that controls the limit cycle of the heart rate. Moreover, the time consequence of CV signals is controlled by the same phase parameter that results in the space dimensionality reduction. These joint equations together with linear assignments to observation are further used in the Kalman filter structure for estimation and tracking. Moreover, we propose a measure of signal fidelity by monitoring the covariance matrix of the innovation signals throughout the filtering procedure. Five categories of life-threatening arrhythmias were verified by simultaneously tracking the signal fidelity and the polar representation of the CV signal estimations. We analyzed data from Physiobank multiparameter databases (MIMIC I and II). Performance evaluation results demonstrated that the sensitivity of the detection ranges over 93.50% and 100.00%. In particular, the addition of more CV signals improved the positive predictivity of the proposed method to 99.27% for the total arrhythmic types. The method was also used for false arrhythmia suppression issued by ICU monitors, with an overall false suppression rate reduced from 42.3% to 9.9%. In addition, false critical ECG arrhythmia alarm rates were found to be, on average, 42.3%, with individual rates varying between 16.7% and 86.5%. The results illustrate that the method can contribute to, and enhance the performance of clinical life-threatening arrhythmia detection.

Journey on Image Clustering Based on Color Composition

Image clustering is a process of grouping images based on their similarity. The image clustering usually uses the color component, texture, edge, shape, or mixture of two components, etc. This research aims to explore image clustering using color composition. In order to complete this image clustering, three main components should be considered, which are color space, image representation (feature extraction), and clustering method itself. We aim to explore which composition of these factors will produce the best clustering results by combining various techniques from the three components. The color spaces use RGB, HSV, and L*a*b* method. The image representations use Histogram and Gaussian Mixture Model (GMM), whereas the clustering methods use K-Means and Agglomerative Hierarchical Clustering algorithm. The results of the experiment show that GMM representation is better combined with RGB and L*a*b* color space, whereas Histogram is better combined with HSV. The experiments also show that K-Means is better than Agglomerative Hierarchical for images clustering. Keywords—Image clustering, feature extraction, RGB, HSV, L*a*b*, Gaussian Mixture Model (GMM), histogram, Agglomerative Hierarchical Clustering (AHC), K-Means, Expectation-Maximization (EM).

Automatic identification of Mycobacterium tuberculosis by Gaussian mixture models

Tuberculosis and other kinds of mycobacteriosis are serious illnesses for which early diagnosis is critical for disease control. Sputum sample analysis is a common manual technique employed for bacillus detection but current sample-analysis techniques are time-consuming, very tedious, subject to poor specificity and require highly trained personnel. Image-processing and pattern-recognition techniques are appropriate tools for improving the manual screening of samples. Here we present a new technique for sputum image analysis that combines invariant shape features and chromatic channel thresholding. Some feature descriptors were extracted from an edited bacillus data set to characterize their shape. They were statistically represented by using a Gaussian mixture model representation and a minimal error Bayesian classification procedure was employed for the last identification stage. This technique constitutes a step towards automating the process and providing a high specificity.