Search

The diagnosis and treatment of individuals with communication disorders offers many opportunities for the application of speech technology, but research so far has not adequately considered: the diversity of conditions, the role of pragmatic deficits, and the challenges of limited data. This paper explores how a general-purpose model of perceived pragmatic similarity may overcome these limitations. It explains how it might support several use cases for clinicians and clients, and presents evidence that a simple model can provide value, and in particular can capture utterance aspects that are relevant to diagnoses of autism and specific language impairment., Comment: submitted to IEEE ICASSP 2025

We investigate which prosodic features matter most in conveying prosodic functions. We use the problem of predicting human perceptions of pragmatic similarity among utterance pairs to evaluate the utility of prosodic features of different types. We find, for example, that duration-related features are more important than pitch-related features, and that utterance-initial features are more important than utterance-final features. Further, failure analysis indicates that modeling using pitch features only often fails to handle important pragmatic functions, and suggests that several generally-neglected acoustic and prosodic features are pragmatically significant, including nasality and vibrato. These findings can guide future basic research in prosody, and suggest how to improve speech synthesis evaluation, among other applications., Comment: Submitted to ICASSP 2025. Audio illustrations available at https://www.cs.utep.edu/nigel/pros-prag/

Large-scale study of glaciers improves our understanding of global glacier change and is imperative for monitoring the ecological environment, preventing disasters, and studying the effects of global climate change. Glaciers in the Hindu Kush Himalaya (HKH) are particularly interesting as the HKH is one of the world's most sensitive regions for climate change. In this work, we: (1) propose a modified version of the U-Net for large-scale, spatially non-overlapping, clean glacial ice, and debris-covered glacial ice segmentation; (2) introduce a novel self-learning boundary-aware loss to improve debris-covered glacial ice segmentation performance; and (3) propose a feature-wise saliency score to understand the contribution of each feature in the multispectral Landsat 7 imagery for glacier mapping. Our results show that the debris-covered glacial ice segmentation model trained using self-learning boundary-aware loss outperformed the model trained using dice loss. Furthermore, we conclude that red, shortwave infrared, and near-infrared bands have the highest contribution toward debris-covered glacial ice segmentation from Landsat 7 images., Comment: Vol. 4 (2023): Proceedings of the Northern Lights Deep Learning Workshop 2023

Normalization layers have been shown to improve convergence in deep neural networks, and even add useful inductive biases. In many vision applications the local spatial context of the features is important, but most common normalization schemes including Group Normalization (GN), Instance Normalization (IN), and Layer Normalization (LN) normalize over the entire spatial dimension of a feature. This can wash out important signals and degrade performance. For example, in applications that use satellite imagery, input images can be arbitrarily large; consequently, it is nonsensical to normalize over the entire area. Positional Normalization (PN), on the other hand, only normalizes over a single spatial position at a time. A natural compromise is to normalize features by local context, while also taking into account group level information. In this paper, we propose Local Context Normalization (LCN): a normalization layer where every feature is normalized based on a window around it and the filters in its group. We propose an algorithmic solution to make LCN efficient for arbitrary window sizes, even if every point in the image has a unique window. LCN outperforms its Batch Normalization (BN), GN, IN, and LN counterparts for object detection, semantic segmentation, and instance segmentation applications in several benchmark datasets, while keeping performance independent of the batch size and facilitating transfer learning., Comment: Accepted as a CVPR 2020 oral paper. arXiv admin note: text overlap with arXiv:1803.08494 by other authors

We focus on the automatic 3D terrain segmentation problem using hyperspectral shortwave IR (HS-SWIR) imagery and 3D Digital Elevation Models (DEM). The datasets were independently collected, and metadata for the HS-SWIR dataset are unavailable. We explore an overall slope of the SWIR spectrum that correlates with the presence of moisture in soil to propose a band ratio test to be used as a proxy for soil moisture content to distinguish two broad classes of objects: live vegetation from impermeable manmade surface. We show that image based localization techniques combined with the Optimal Randomized RANdom Sample Consensus (RANSAC) algorithm achieve precise spatial matches between HS-SWIR data of a portion of downtown Los Angeles (LA (USA)) and the Visible image of a geo-registered 3D DEM, covering a wider-area of LA. Our spectral-elevation rule based approach yields an overall accuracy of 97.7%, segmenting the object classes into buildings, houses, trees, grass, and roads/parking lots., Comment: Published on: IEEE Workshop on Hyperspectral Image and Signal Processing Conference (WHISPERS 2018), Amsterdam, The Netherlands, September 2018

The authors of this report met on 28-30 March 2018 at the New Jersey Institute of Technology, Newark, New Jersey, for a 3-day workshop that brought together a group of data providers, expert modelers, and computer and data scientists, in the solar discipline. Their objective was to identify challenges in the path towards building an effective framework to achieve transformative advances in the understanding and forecasting of the Sun-Earth system from the upper convection zone of the Sun to the Earth's magnetosphere. The workshop aimed to develop a research roadmap that targets the scientific challenge of coupling observations and modeling with emerging data-science research to extract knowledge from the large volumes of data (observed and simulated) while stimulating computer science with new research applications. The desire among the attendees was to promote future trans-disciplinary collaborations and identify areas of convergence across disciplines. The workshop combined a set of plenary sessions featuring invited introductory talks and workshop progress reports, interleaved with a set of breakout sessions focused on specific topics of interest. Each breakout group generated short documents, listing the challenges identified during their discussions in addition to possible ways of attacking them collectively. These documents were combined into this report-wherein a list of prioritized activities have been collated, shared and endorsed., Comment: Workshop Report

We have developed a method for fast and accurate stellar population parameters determination in order to apply it to high resolution galaxy spectra. The method is based on an optimization technique that combines active learning with an instance-based machine learning algorithm. We tested the method with the retrieval of the star-formation history and dust content in "synthetic" galaxies with a wide range of S/N ratios. The "synthetic" galaxies where constructed using two different grids of high resolution theoretical population synthesis models. The results of our controlled experiment shows that our method can estimate with good speed and accuracy the parameters of the stellar populations that make up the galaxy even for very low S/N input. For a spectrum with S/N=5 the typical average deviation between the input and fitted spectrum is less than 10**{-5}. Additional improvements are achieved using prior knowledge., Comment: 14 pages, 25 figures, accepted by Monthly Notices

In this work we focus on the determination of the relative distributions of young, intermediate-age and old populations of stars in galaxies. Starting from a grid of theoretical population synthesis models we constructed a set of model galaxies with a distribution of ages, metallicities and intrinsic reddening. Using this set we have explored a new fitting method that presents several advantages over conventional methods. We propose an optimization technique that combines active learning with an instance-based machine learning algorithm. Experimental results show that this method can estimate with high speed and accuracy the physical parameters of the stellar populations., Comment: 4 pages, 1 figure, To appear in Proceedings of ADASS XIII, Strasbourg, October 2003

We present a new technique to segregate old and young stellar populations in galactic spectra using machine learning methods. We used an ensemble of classifiers, each classifier in the ensemble specializes in young or old populations and was trained with locally weighted regression and tested using ten-fold cross-validation. Since the relevant information concentrates in certain regions of the spectra we used the method of sequential floating backward selection offline for feature selection. The application to Seyfert galaxies proved that this technique is very insensitive to the dilution by the Active Galactic Nucleus (AGN) continuum. Comparing with exhaustive search we concluded that both methods are similar in terms of accuracy but the machine learning method is faster by about two orders of magnitude., Comment: 4 pages, 1 figure, To appear in Proceedings of the ADASS-XIII Conference in Strasbourg, October 2003