Multi-resolution Model Research Articles

Predicting speech intelligibility with deep neural networks

An accurate objective prediction of human speech intelligibility is of interest for many applications such as the evaluation of signal processing algorithms. To predict the speech recognition threshold (SRT) of normal-hearing listeners, an automatic speech recognition (ASR) system is employed that uses a deep neural network (DNN) to convert the acoustic input into phoneme predictions, which are subsequently decoded into word transcripts. ASR results are obtained with and compared to data presented in Schubotz et al. (2016), which comprises eight different additive maskers that range from speech-shaped stationary noise to a single-talker interferer and responses from eight normal-hearing subjects. The task for listeners and ASR is to identify noisy words from a German matrix sentence test in monaural conditions. Two ASR training schemes typically used in applications are considered: (A) matched training, which uses the same noise type for training and testing and (B) multi-condition training, which covers all eight maskers. For both training schemes, ASR-based predictions outperform established measures such as the extended speech intelligibility index (ESII), the multi-resolution speech envelope power spectrum model (mr-sEPSM) and others. This result is obtained with a speaker-independent model that compares the word labels of the utterance with the ASR transcript, which does not require separate noise and speech signals. The best predictions are obtained for multi-condition training with amplitude modulation features, which implies that the noise type has been seen during training. Predictions and measurements are analyzed by comparing speech recognition thresholds and individual psychometric functions to the DNN-based results.

Multi-resolution geometric modeling of the mitral heart valve leaflets.

An essential element of cardiac function, the mitral valve (MV) ensures proper directional blood flow between the left heart chambers. Over the past two decades, computational simulations have made marked advancements toward providing powerful predictive tools to better understand valvular function and improve treatments for MV disease. However, challenges remain in the development of robust means for the quantification and representation of MV leaflet geometry. In this study, we present a novel modeling pipeline to quantitatively characterize and represent MV leaflet surface geometry. Our methodology utilized a two-part additive decomposition of the MV geometric features to decouple the macro-level general leaflet shape descriptors from the leaflet fine-scale features. First, the general shapes of five ovine MV leaflets were modeled using superquadric surfaces. Second, the finer-scale geometric details were captured, quantified, and reconstructed via a 2D Fourier analysis with an additional sparsity constraint. This spectral approach allowed us to easily control the level of geometric details in the reconstructed geometry. The results revealed that our methodology provided a robust and accurate approach to develop MV-specific models with an adjustable level of spatial resolution and geometric detail. Such fully customizable models provide the necessary means to perform computational simulations of the MV at a range of geometric accuracies in order to identify the level of complexity required to achieve predictive MV simulations.

GENERATION ALGORITHM OF DISCRETE LINE IN MULTI-DIMENSIONAL GRIDS

Abstract. Discrete Global Grids System (DGGS) is a kind of digital multi-resolution earth reference model, in terms of structure, it is conducive to the geographical spatial big data integration and mining. Vector is one of the important types of spatial data, only by discretization, can it be applied in grids system to make process and analysis. Based on the some constraint conditions, this paper put forward a strict definition of discrete lines, building a mathematic model of the discrete lines by base vectors combination method. Transforming mesh discrete lines issue in n-dimensional grids into the issue of optimal deviated path in n-minus-one dimension using hyperplane, which, therefore realizing dimension reduction process in the expression of mesh discrete lines. On this basis, we designed a simple and efficient algorithm for dimension reduction and generation of the discrete lines. The experimental results show that our algorithm not only can be applied in the two-dimensional rectangular grid, also can be applied in the two-dimensional hexagonal grid and the three-dimensional cubic grid. Meanwhile, when our algorithm is applied in two-dimensional rectangular grid, it can get a discrete line which is more similar to the line in the Euclidean space.

Spatiotemporal multiresolution modeling to infill missing areal data and enhance the temporal frequency of infrared satellite images

High‐resolution environmental observations from satellite remote sensing are useful for gaining insight into climate variability over large regions of the earth. However, data products derived from infrared remote sensing can have large amounts of missing data due to periodic cloud cover obscuring the earth's surface. In addition, these products are commonly derived from instruments aboard polar‐orbiting satellites, which make relatively infrequent passes over each region of the planet, leaving large blocks of data that are missing in time. The goal of this research is to overcome these issues by spatially and temporally infilling temperature data using satellite observations. To accomplish this, we propose novel statistical modeling techniques to account for dissonance between the areal spatial support of the data and the continuous temporal domain. In addition, we develop a novel multiresolution approach that accommodates the large spatial dimensionality of high‐resolution satellite images. The result of our work is a computationally efficient methodology that provides the means to infill satellite data products both spatially and temporally. The methodology is demonstrated using land surface temperature fields from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument onboard NASA's Terra and Aqua satellites, over Harris County, Texas, USA.

A comprehensive pipeline for multi-resolution modeling of the mitral valve: Validation, computational efficiency, and predictive capability.

Multiple studies have demonstrated that the pathological geometries unique to each patient can affect the durability of mitral valve (MV) repairs. While computational modeling of the MV is a promising approach to improve the surgical outcomes, the complex MV geometry precludes use of simplified models. Moreover, the lack of complete in vivo geometric information presents significant challenges in the development of patient-specific computational models. There is thus a need to determine the level of detail necessary for predictive MV models. To address this issue, we have developed a novel pipeline for building attribute-rich computational models of MV with varying fidelity directly from the in vitro imaging data. The approach combines high-resolution geometric information from loaded and unloaded states to achieve a high level of anatomic detail, followed by mapping and parametric embedding of tissue attributes to build a high-resolution, attribute-rich computational models. Subsequent lower resolution models were then developed and evaluated by comparing the displacements and surface strains to those extracted from the imaging data. We then identified the critical levels of fidelity for building predictive MV models in the dilated and repaired states. We demonstrated that a model with a feature size of about 5mm and mesh size of about 1mm was sufficient to predict the overall MV shape, stress, and strain distributions with high accuracy. However, we also noted that more detailed models were found to be needed to simulate microstructural events. We conclude that the developed pipeline enables sufficiently complex models for biomechanical simulations of MV in normal, dilated, repaired states.

A non-linear and non-stationary perspective for downscaling mean monthly temperature: a wavelet coupled second order Volterra model

This study presents a multiscale framework for downscaling of the General Circulation Model (GCM) outputs to the mean monthly temperature at regional scale using a wavelet based Second order Voltera (SoV) model. The models are developed using the reanalysis climatic data from the National Centers for Environmental Prediction (NCEP) and are validated using the simulated climatic dataset from the Can CM4 GCM for five locations in the Krishna river basin, India. K-means clustering, based on the multiscale wavelet entropy of the predictors, is used for obtaining the clusters of the input climatic variables. Principal component analysis (PCA) is used to obtain the representative variables from each cluster. These input variables are then used to develop a wavelet based multiscale model using Second order Volterra approach to simulate observed mean monthly temperature for the selected locations in the basin. These models are called W-P-SoV models in this paper. For the purpose of comparison, linear multi-resolution models are developed using Multiple Linear regression (MLR) and are called W-P MLR models. The performance of the models is further compared with other Wavelet-PCA based models coupled with Multiple linear regression models (P-MLR) and Artificial Neural Networks (P-ANN), and, stand-alone MLR and ANN to establish the superiority of the proposed approach. The results indicate that the performance of the wavelet based models is superior in terms of downscaling accuracy when compared with the other models used.

Geometric multiscale community detection: Markov stability and vector partitioning

Multiscale community detection can be viewed from a dynamical perspective within the Markov Stability framework, which uses the diffusion of a Markov process on the graph to uncover intrinsic network substructures across all scales. Here we reformulate multiscale community detection as a max-sum length vector partitioning problem with respect to the set of time-dependent node vectors expressed in terms of eigenvectors of the transition matrix. This formulation provides a geometric interpretation of Markov Stability in terms of a time-dependent spectral embedding, where the Markov time acts as an inhomogeneous geometric resolution factor that zooms the components of the node vectors at different rates. Our geometric formulation encompasses both modularity and the multi-resolution Potts model, which are shown to correspond to vector partitioning in a pseudo-Euclidean space, and is also linked to spectral partitioning methods, where the number of eigenvectors used corresponds to the dimensionality of the underlying embedding vector space. Inspired by the Louvain optimisation for community detection, we then propose an algorithm based on a graph-theoretical heuristic for the vector partitioning problem. We apply the algorithm to the spectral optimisation of modularity and Markov Stability community detection. The spectral embedding based on the transition matrix eigenvectors leads to improved partitions with higher information content and higher modularity than the eigen-decomposition of the modularity matrix. We illustrate the results with random network benchmarks.

A general CFD framework for fault-resilient simulations based on multi-resolution information fusion

We develop a general CFD framework for multi-resolution simulations to target multiscale problems but also resilience in exascale simulations, where faulty processors may lead to gappy, in space-time, simulated fields. We combine approximation theory and domain decomposition together with statistical learning techniques, e.g. coKriging, to estimate boundary conditions and minimize communications by performing independent parallel runs. To demonstrate this new simulation approach, we consider two benchmark problems. First, we solve the heat equation (a) on a small number of spatial “patches” distributed across the domain, simulated by finite differences at fine resolution and (b) on the entire domain simulated at very low resolution, thus fusing multi-resolution models to obtain the final answer. Second, we simulate the flow in a lid-driven cavity in an analogous fashion, by fusing finite difference solutions obtained with fine and low resolution assuming gappy data sets. We investigate the influence of various parameters for this framework, including the correlation kernel, the size of a buffer employed in estimating boundary conditions, the coarseness of the resolution of auxiliary data, and the communication frequency across different patches in fusing the information at different resolution levels. In addition to its robustness and resilience, the new framework can be employed to generalize previous multiscale approaches involving heterogeneous discretizations or even fundamentally different flow descriptions, e.g. in continuum-atomistic simulations.

Multi-resolution multi-trial sparse classification model for decoding visual memories from hippocampal spikes in human.

To understand how memories are encoded in the hippocampus, we build memory decoding models to classify visual memories based on hippocampal activities in human. Model inputs are spatio-temporal patterns of spikes recorded in the hippocampal CA3 and CA1 regions of epilepsy patients performing a delayed match-to-sample (DMS) task. Model outputs are binary labels indicating categories and features of sample images. To solve the super high-dimensional estimation problem with short data length, we develop a multi-trial, sparse model estimation method utilizing B-spline basis functions with a large range of temporal resolutions and a regularized logistic classifier. Results show that this model can effectively avoid overfitting and provide significant amount of prediction to memory categories and features using very limited number of data points. Stable estimation of sparse classification function matrices for each label can be obtained with this multi-resolution, multi-trial procedure. These classification models can be used not only to predict memory contents, but also to design optimal spatio-temporal patterns for eliciting specific memories in the hippocampus, and thus have important implications to the development of hippocampal memory prostheses.

Constructing and evaluating multi-resolution model pairs: an attrition modeling example

Our aim is to illustrate how to apply the methodology for characterizing the useful approximation zone of a lumping process, as described in an earlier note in this journal. Here we review the procedure steps and then go on to present an example in the combat modeling area. We conclude by discussing implications for approximate morphism support of multi-resolution modeling.

Obstacle avoidance behavior of swarm robots based on aggregation and disaggregation method

One of the current challenges in the development of swarm robots is making them realize obstacle avoidance while keeping formation. This paper proposes a new obstacle avoidance method for multiple robots based on the aggregation and disaggregation approach. First, the formation-based multi-resolution models of swarm robots are presented and the formation attribute-based consistency mapping functions are designed. Then, obstacle avoidance behavior based on aggregation and disaggregation method is presented and the basic obstacle avoidance behaviors are designed. Finally, the simulation results are presented to show the feasibility of the proposed method.

A morphologically preserved multi-resolution TIN surface modeling and visualization method for virtual globes

Virtual globes play an important role in representing three-dimensional models of the Earth. To extend the functioning of a virtual globe beyond that of a “geobrowser”, the accuracy of the geospatial data in the processing and representation should be of special concern for the scientific analysis and evaluation. In this study, we propose a method for the processing of large-scale terrain data for virtual globe visualization and analysis. The proposed method aims to construct a morphologically preserved multi-resolution triangulated irregular network (TIN) pyramid for virtual globes to accurately represent the landscape surface and simultaneously satisfy the demands of applications at different scales. By introducing cartographic principles, the TIN model in each layer is controlled with a data quality standard to formulize its level of detail generation. A point-additive algorithm is used to iteratively construct the multi-resolution TIN pyramid. The extracted landscape features are also incorporated to constrain the TIN structure, thus preserving the basic morphological shapes of the terrain surface at different levels. During the iterative construction process, the TIN in each layer is seamlessly partitioned based on a virtual node structure, and tiled with a global quadtree structure. Finally, an adaptive tessellation approach is adopted to eliminate terrain cracks in the real-time out-of-core spherical terrain rendering. The experiments undertaken in this study confirmed that the proposed method performs well in multi-resolution terrain representation, and produces high-quality underlying data that satisfy the demands of scientific analysis and evaluation.

Sharp volumetric billboard based characterization and modeling of complex 3D Ni/Al high energy ball milled composites

We present an innovative image-based modeling technique, based on Google Earth like algorithms, to effectively resolve intricate material morphology and address the computational complexity associated with heterogeneous materials. This sharp volumetric billboard algorithm stems from a volumetric billboard method, a multi-resolution modeling strategy in computer graphics. In this work, we enhance volumetric billboards through a sharpening filter to reconstruct the statistical information of heterogeneous systems. A hierarchy of microstructures is created for high energy ball milled Ni/Al composites. We analyze the first- and second-order statistics of microstructures, and characterize both macro- and micro-mechanical material responses. Furthermore, we conduct a convergence study of the associated computational results. The statistical and mechanical robustness of data compression is demonstrated through the corresponding error analysis.

Multi-resolution model of an industrial hydrogen plant for plantwide operational optimization with non-uniform steam-methane reformer temperature field

Hydrogen is consumed in large quantities in the chemical industry. The most common industrial process for hydrogen production is steam-methane reforming, which is carried out using an energy-intensive furnace. The plant energy efficiency depends strongly on the spatial temperature distribution within the furnace; the narrower the distribution, the higher the efficiency that can be achieved. However, currently available studies on plantwide optimization of hydrogen plants ignore this crucial aspect. Adequate resolution of the spatial temperature distribution is necessary to determine the furnace operating temperature, which, in turn, determines the plant efficiency. In this work, a multi-resolution model of a hydrogen plant is developed. It includes a high-resolution model of the furnace, and low-resolution models, adequate for the purpose of plantwide optimization, of other unit operations. The developed model is used to determine the optimal process conditions after furnace temperature homogenization as part of a plant start-up or setpoint changeover procedure.

Multi-resolution multi-object statistical shape models based on the locality assumption.

Statistical shape models learned from a population of previously observed training shapes are nowadays widely used in medical image analysis to aid segmentation or classification. However, providing an appropriate and representative training population of preferably manual segmentations is typically either very labor-intensive or even impossible. Therefore, statistical shape models in practice frequently suffer from the high-dimension-low-sample-size (HDLSS) problem resulting in models with insufficient expressiveness. In this paper, a novel approach for learning representative multi-resolution multi-object statistical shape models from a small number of training samples that adequately model the variability of each individual object as well as their interrelations is presented. The method is based on the assumption of locality, which means that local shape variations have limited effects in distant areas and, therefore, can be modeled independently. This locality assumption is integrated into the standard statistical shape modeling framework by manipulating the sample covariance matrix (non-zero covariances between distant landmarks are set to zero). To allow for multi-object modeling, a method for computing distances between points located on different object shapes is proposed. Furthermore, different levels of locality are introduced by deriving a multi-resolution scheme, which is equipped with a method to combine variability information modeled at different levels into a single shape model. This combined representation of global and local variability in a single shape model allows the use of the classical active shape model strategy for model-based image segmentation. An extensive evaluation based on a public data base of 247 chest radiographs is performed to show the modeling and segmentation capabilities of the proposed approach in single- and multi-object HDLSS scenarios. The new approach is not only compared to the classical shape modeling method but also to three state-of-the-art shape modeling approaches specifically designed to cope with the HDLSS problem. The results show that the new approach significantly outperforms all other approaches in terms of generalization ability and model-based segmentation accuracy.

A high-fidelity multiresolution digital elevation model for Earth systems

Abstract. The impact of topography on Earth systems variability is well recognised. As numerical simulations evolved to incorporate broader scales and finer processes, accurately assimilating or transforming the topography to produce more exact land–atmosphere–ocean interactions, has proven to be quite challenging. Numerical schemes of Earth systems often use empirical parameterisation at sub-grid scale with downscaling to express topographic endogenous processes, or rely on insecure point interpolation to induce topographic forcing, which creates bias and input uncertainties. Digital elevation model (DEM) generalisation provides more sophisticated systematic topographic transformation, but existing methods are often difficult to be incorporated because of unwarranted grid quality. Meanwhile, approaches over discrete sets often employ heuristic approximation, which are generally not best performed. Based on DEM generalisation, this article proposes a high-fidelity multiresolution DEM with guaranteed grid quality for Earth systems. The generalised DEM surface is initially approximated as a triangulated irregular network (TIN) via selected feature points and possible input features. The TIN surface is then optimised through an energy-minimised centroidal Voronoi tessellation (CVT). By devising a robust discrete curvature as density function and exact geometry clipping as energy reference, the developed curvature CVT (cCVT) converges, the generalised surface evolves to a further approximation to the original DEM surface, and the points with the dual triangles become spatially equalised with the curvature distribution, exhibiting a quasi-uniform high-quality and adaptive variable resolution. The cCVT model was then evaluated on real lidar-derived DEM datasets and compared to the classical heuristic model. The experimental results show that the cCVT multiresolution model outperforms classical heuristic DEM generalisations in terms of both surface approximation precision and surface morphology retention.

Multi-time-scale hydroclimate dynamics of a regional watershed and links to large-scale atmospheric circulation: Application to the Seine river catchment, France

In the present context of global changes, considerable efforts have been deployed by the hydrological scientific community to improve our understanding of the impacts of climate fluctuations on water resources. Both observational and modeling studies have been extensively employed to characterize hydrological changes and trends, assess the impact of climate variability or provide future scenarios of water resources. In the aim of a better understanding of hydrological changes, it is of crucial importance to determine how and to what extent trends and long-term oscillations detectable in hydrological variables are linked to global climate oscillations.In this work, we develop an approach associating correlation between large and local scales, empirical statistical downscaling and wavelet multiresolution decomposition of monthly precipitation and streamflow over the Seine river watershed, and the North Atlantic sea level pressure (SLP) in order to gain additional insights on the atmospheric patterns associated with the regional hydrology. We hypothesized that: (i) atmospheric patterns may change according to the different temporal wavelengths defining the variability of the signals; and (ii) definition of those hydrological/circulation relationships for each temporal wavelength may improve the determination of large-scale predictors of local variations.The results showed that the links between large and local scales were not necessarily constant according to time-scale (i.e. for the different frequencies characterizing the signals), resulting in changing spatial patterns across scales. This was then taken into account by developing an empirical statistical downscaling (ESD) modeling approach, which integrated discrete wavelet multiresolution analysis for reconstructing monthly regional hydrometeorological processes (predictand: precipitation and streamflow on the Seine river catchment) based on a large-scale predictor (SLP over the Euro-Atlantic sector). This approach basically consisted in three steps: 1 – decomposing large-scale climate and hydrological signals (SLP field, precipitation or streamflow) using discrete wavelet multiresolution analysis, 2 – generating a statistical downscaling model per time-scale, 3 – summing up all scale-dependent models in order to obtain a final reconstruction of the predictand. The results obtained revealed a significant improvement of the reconstructions for both precipitation and streamflow when using the multiresolution ESD model instead of basic ESD. In particular, the multiresolution ESD model handled very well the significant changes in variance through time observed in either precipitation or streamflow. For instance, the post-1980 period, which had been characterized by particularly high amplitudes in interannual-to-interdecadal variability associated with alternating flood and extremely low-flow/drought periods (e.g., winter/spring 2001, summer 2003), could not be reconstructed without integrating wavelet multiresolution analysis into the model. In accordance with previous studies, the wavelet components detected in SLP, precipitation and streamflow on interannual to interdecadal time-scales could be interpreted in terms of influence of the Gulf-Stream oceanic front on atmospheric circulation.

Multi-Resolution Feature Fusion model for coal rock burst hazard recognition based on Acoustic Emission data

Abstract The Acoustic Emission (AE) technique is an important nondestructive testing method used to detect coal rock burst hazards in coal mine strata to ensure the safety of the lives of miners. It is, however, still difficult to accurately detect coal rock bursts due to the complex mechanism of AE propagation underground. Therefore, this study proposes a new Multi-Resolution Feature Fusion SVM recognition (MRFF-SVM) approach to compute a comprehensive feature vector for coal rock burst hazard recognition and forecasting. The proposed approach contains the following three improved processes: the Coiflet Wavelet Transform (CWT) to decompose AE waveforms into multiple perspectives for feature vectors extraction, the Multi-Resolution Feature Fusion (MRFF) method to fuse these selected feature vectors into an enhanced MRFF feature vector and the following Support Vector Machines (SVM) to recognize the coal rock burst conditions. Several innovative experiments are carried out to evaluate the coal rock conditions according to the safe, crisis and burst categories. The results indicate that the proposed MRFF feature vectors can retain AE signatures well to recognize coal rock conditions and forecast sandstone failure. Thus, MRFF-SVM provides an effective analysis tool for coal rock burst hazard detection.

Traffic Multiresolution Modeling and Consistency Analysis of Urban Expressway Based on Asynchronous Integration Strategy

The paper studies multiresolution traffic flow simulation model of urban expressway. Firstly, compared with two-level hybrid model, three-level multiresolution hybrid model has been chosen. Then, multiresolution simulation framework and integration strategies are introduced. Thirdly, the paper proposes an urban expressway multiresolution traffic simulation model by asynchronous integration strategy based on Set Theory, which includes three submodels: macromodel, mesomodel, and micromodel. After that, the applicable conditions and derivation process of the three submodels are discussed in detail. In addition, in order to simulate and evaluate the multiresolution model, “simple simulation scenario” of North-South Elevated Expressway in Shanghai has been established. The simulation results showed the following.(1)Volume-density relationships of three submodels are unanimous with detector data.(2)When traffic density is high, macromodel has a high precision and smaller error and the dispersion of results is smaller. Compared with macromodel, simulation accuracies of micromodel and mesomodel are lower but errors are bigger.(3)Multiresolution model can simulate characteristics of traffic flow, capture traffic wave, and keep the consistency of traffic state transition. Finally, the results showed that the novel multiresolution model can have higher simulation accuracy and it is feasible and effective in the real traffic simulation scenario.

Increase the capacity of distributed energy resource by multi‐dimensional multi‐resolution model

In order to deal with the problem of energy and environment, it is important to increase the maximum capacity of distributed energy resource (DER) in the distribution network. A multi-dimensional multi-resolution model (MDMRM) of distribution network and a method are proposed to calculate and increase the capacity of DER considering voltage, power loss, and thermal constraints. Finally, the model and method are applied to the IEEE Comprehensive Case. The simulation results show that the calculations and analysis of DER capacity have been proved to be less time-consuming based on the MDMRM, and the proposed method is a useful solution to increase DER capacity. It is important for reliable, high-quality distribution network with large capacity of DER.