Visar Berisha

Sparse representations for automatic target classification in SAR images

We propose a sparse representation approach for classifying different targets in Synthetic Aperture Radar (SAR) images. Unlike the other feature based approaches, the proposed method does not require explicit pose estimation or any preprocessing. The dictionary used in this setup is the collection of the normalized training vectors itself. Computing a sparse representation for the test data using this dictionary corresponds to finding a locally linear approximation with respect to the underlying class manifold. SAR images obtained from the Moving and Stationary Target Acquisition and Recognition (MSTAR) public database were used in the classification setup. Results show that the performance of the algorithm is superior to using a support vector machines based approach with similar assumptions. Significant complexity reduction is obtained by reducing the dimensions of the data using random projections for only a small loss in performance.

Tracking Discourse Complexity Preceding Alzheimer's Disease Diagnosis: A Case Study Comparing the Press Conferences of Presidents Ronald Reagan and George Herbert Walker Bush

Changes in some lexical features of language have been associated with the onset and progression of Alzheimers disease. Here we describe a method to extract key features from discourse transcripts, which we evaluated on non-scripted news conferences from President Ronald Reagan, who was diagnosed with Alzheimers disease in 1994, and President George Herbert Walker Bush, who has no known diagnosis of Alzheimers disease. Key word counts previously associated with cognitive decline in Alzheimers disease were extracted and regression analyses were conducted. President Reagan showed a significant reduction in the number of unique words over time and a significant increase in conversational fillers and non-specific nouns over time. There was no significant trend in these features for President Bush.

Automatic assessment of vowel space area.

Vowel space area (VSA) is an attractive metric for the study of speech production deficits and reductions in intelligibility, in addition to the traditional study of vowel distinctiveness. Traditional VSA estimates are not currently sufficiently sensitive to map to production deficits. The present report describes an automated algorithm using healthy, connected speech rather than single syllables and estimates the entire vowel working space rather than corner vowels. Analyses reveal a strong correlation between the traditional VSA and automated estimates. When the two methods diverge, the automated method seems to provide a more accurate area since it accounts for all vowels.

Empirically Estimable Classification Bounds Based on a Nonparametric Divergence Measure

Information divergence functions play a critical role in statistics and information theory. In this paper we show that a nonparametric f-divergence measure can be used to provide improved bounds on the minimum binary classification probability of error for the case when the training and test data are drawn from the same distribution and for the case where there exists some mismatch between training and test distributions. We confirm these theoretical results by designing feature selection algorithms using the criteria from these bounds and by evaluating the algorithms on a series of pathological speech classification tasks.

Experiments With Sensor Motes and Java-DSP

Distributed wireless sensor networks (WSNs) are being proposed for various applications including defense, security, and smart stages. The introduction of hardware wireless sensors in a signal processing education setting can serve as a paradigm for data acquisition, collaborative signal processing, or simply as a platform for obtaining, processing, and analyzing real-life real-time data. In this paper, a software interface that enables the Java-digital signal processing (J-DSP) visual programming environment to communicate in a two-way manner with a wireless sensor network is presented. This interface was developed by writing nesC (an extension to the C programming language for sensors) code that enables J-DSP to issue commands to multiple wireless sensor motes, activate specific transducers, and analyze data using any of the existing J-DSP signal processing functions in real time. A series of exercises were developed and disseminated to provide hardware experiences to signals and systems and signal processing undergraduate students. The hardware with the J-DSP software has been used for two semesters in the senior level digital signal processing (DSP) course at Arizona State University. The interface, the exercises, and their assessment (instruments and results) are described in the paper.

Real-time acoustic monitoring using wireless sensor motes

Wireless sensor networks (WSN) have recently gained popularity in distributed monitoring and surveillance applications. The objective of these devices is to extract pertinent information under several constrains such as low computational capabilities, limited arithmetic precision, and the need to conserve power. One of the most revealing environmental cues is audio. In this paper, we propose a voice activity detector and a simple gender classifier for use in a distributed acoustic sensing system. This algorithm makes use of low-complexity audio features and a pre-trained regression tree to classify incoming speech by gender. The algorithm is implemented real-time on the Crossbow sensor motes and a series of results are given that characterize the algorithm performance and complexity. Challenges in this real-time implementation include designing the algorithm and software architecture such that the signal processing is appropriately distributed between the sensor mote and the base station. At the base station, a data fusion algorithm considers a linear combination of individual mote decisions to form a final decision

Temporal Lobe Cortical Thickness Correlations Differentiate the Migraine Brain from the Healthy Brain

Background Interregional cortical thickness correlations reflect underlying brain structural connectivity and functional connectivity. A few prior studies have shown that migraine is associated with atypical cortical brain structure and atypical functional connectivity amongst cortical regions that participate in sensory processing. However, the specific brain regions that most accurately differentiate the migraine brain from the healthy brain have yet to be determined. The aim of this study was to identify the brain regions that comprised interregional cortical thickness correlations that most differed between migraineurs and healthy controls. Methods This was a cross-sectional brain magnetic resonance imaging (MRI) investigation of 64 adults with migraine and 39 healthy control subjects recruited from tertiary-care medical centers and their surrounding communities. All subjects underwent structural brain MRI imaging on a 3T scanner. Cortical thickness was determined for 70 brain regions that cover the cerebral cortex and cortical thickness correlations amongst these regions were calculated. Cortical thickness correlations that best differentiated groups of six migraineurs from controls and vice versa were identified. Results A model containing 15 interregional cortical thickness correlations differentiated groups of migraineurs from healthy controls with high accuracy. The right temporal pole was involved in 13 of the 15 interregional correlations while the right middle temporal cortex was involved in the other two. Conclusions A model consisting of 15 interregional cortical thickness correlations accurately differentiates the brains of small groups of migraineurs from those of healthy controls. Correlations with the right temporal pole were highly represented in this classifier, suggesting that this region plays an important role in migraine pathophysiology.

Empirical Non-Parametric Estimation of the Fisher Information

The Fisher information matrix (FIM) is a foundational concept in statistical signal processing. The FIM depends on the probability distribution, assumed to belong to a smooth parametric family. Traditional approaches to estimating the FIM require estimating the probability distribution function (PDF), or its parameters, along with its gradient or Hessian. However, in many practical situations the PDF of the data is not known but the statistician has access to an observation sample for any parameter value. Here we propose a method of estimating the FIM directly from sampled data that does not require knowledge of the underlying PDF. The method is based on non-parametric estimation of an f-divergence over a local neighborhood of the parameter space and a relation between curvature of the f-divergence and the FIM. Thus we obtain an empirical estimator of the FIM that does not require density estimation and is asymptotically consistent. We empirically evaluate the validity of our approach using two experiments.

Real-Time Implementation of a Distributed Voice Activity Detector

Wireless sensor networks have been applied successfully in real-time distributed and collaborative sensing. In these situations, each sensor is responsible for extracting pertinent information from the surrounding environment and transmitting it to other sensors and/or to the main processing station. This is done while operating under several constraints, such as low computational capabilities, limited arithmetic precision, and the need to conserve power. In this paper, we present a low-complexity voice activity detector and a gender classifier for implementation on the Crossbow sensor motes. In addition, a decision fusion algorithm that resides at the base station is also implemented. A series of experiments that characterize the performance of the algorithms under varying conditions and in different environments are presented and several of the challenges we faced in developing this real-time implementation are discussed

Modeling pathological speech perception from data with similarity labels

The current state of the art in judging pathological speech intelligibility is subjective assessment performed by trained speech pathologists (SLP). These tests, however, are inconsistent, costly and, oftentimes suffer from poor intra- and inter-judge reliability. As such, consistent, reliable, and perceptually-relevant objective evaluations of pathological speech are critical. Here, we propose a data-driven approach to this problem. We propose new cost functions for examining data from a series of experiments, whereby we ask certified SLPs to rate pathological speech along the perceptual dimensions that contribute to decreased intelligibility. We consider qualitative feedback from SLPs in the form of comparisons similar to statements “Is Speaker As rhythm more similar to Speaker B or Speaker C?” Data of this form is common in behavioral research, but is different from the traditional data structures expected in supervised (data matrix + class labels) or unsupervised (data matrix) machine learning. The proposed method identifies relevant acoustic features that correlate with the ordinal data collected during the experiment. Using these features, we show that we are able to develop objective measures of the speech signal degradation that correlate well with SLP responses.