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

Purpose: Automatic measurements of fundamental frequency ( F 0) typically contain tracking errors that can be challenging to accurately correct. This study assessed what degree these change 0 summary statistics in speakers with Parkinson's disease (PD) and neurotypical adults. In addition, we include a case examining how the removal influenced our ability predict perceptual outcome measure, speech expressiveness, associated dysarthria PD. Several different statistical approaches for...

State-of-the-art automatic speech recognition (ASR) engines perform well on healthy speech; however recent studies show that their performance dysarthric is highly variable. This because of the acoustic variability associated with different dysarthria subtypes. paper aims to develop a better understanding how perceptual disturbances in relate ASR performance. Accurate ratings representative set 32 speakers along dimensions are obtained and algorithm same analyzed. work explores relationship...

Direct decoding of speech from the brain is a faster alternative to current electroencephalography (EEG) speller-based brain-computer interfaces (BCI) in providing communication assistance locked-in patients. Magnetoencephalography (MEG) has recently shown great potential as non-invasive neuroimaging modality for neural decoding, owing part its spatial selectivity over other high-temporal resolution devices. Standard MEG systems have large number cryogenically cooled channels/sensors (200 -...

The spatiotemporal index (STI) is a widely used approach for measuring speech pattern stability across multiple repetitions of stimulus. In this study, we examine how methodological choices in the implementation STI (including number repetitions, length stimuli, and parsing procedure) can affect its value.

Silent speech interfaces (SSIs) convert non-audio bio-signals, such as articulatory movement, to speech. This technology has the potential recover ability of individuals who have lost their voice but can still articulate (e.g., laryngectomees). Articulation-to-speech (ATS) synthesis is an algorithm design SSI that advantages easy-implementation and low-latency, therefore becoming more popular. Current ATS studies focus on speaker-dependent (SD) models avoid large variations patterns acoustic...

This study investigated whether listener processing of dysarthric speech requires the recruitment more cognitive resources (i.e., higher levels listening effort) than neurotypical speech. We also explored relationships between behavioral effort, perceived and objective measures word transcription accuracy.A recall paradigm was used to index effort. The primary task involved transcription, whereas a memory recalling words from previous sentences. Nineteen listeners completed twice, once while...

Behavioral speech modifications have variable effects on the intelligibility of speakers with dysarthria. In companion article, a significant relationship was found between measures speakers' baseline and their gains following cues to speak louder reduce rate (Fletcher, McAuliffe, Lansford, Sinex, & Liss, 2017). This study reexamines these features assesses whether automated acoustic assessments can also be used predict gains.Fifty (7 older individuals 43 dysarthria) read passage in...

A number of fundamental quantities in statistical signal processing and information theory can be expressed as integral functions two probability density functions. Such are called functionals they map onto the real line. For example, divergence measure dissimilarity between useful a applications. Typically, estimating these requires complete knowledge underlying distribution followed by multi-dimensional integration. Existing methods make parametric assumptions about data or use...

Existing speech classification algorithms often perform well when evaluated on training and test data drawn from the same distribution. In practice, however, these distributions are not always same. circumstances, performance of trained models will likely decrease. this paper, we discuss an underutilized divergence measure derive estimable upper bound error rate that depends distance between distributions. Using as motivation, develop a feature learning algorithm aims to identify invariant...

The spatiotemporal index (STI) is a standard metric for quantifying the stability and patterning of speech movements. STI has often been applied to individual articulators, but an derived from acoustic signal offers composite easily obtained measure that incorporates multiple components production complex. In this work, we examine relationship between kinematic STIs in children with without developmental language disorder (DLD), aim determining whether reflect similar degrees variability.

In this paper, we extend previously developed non-parametric bounds on the Bayes risk in binary classification problems to multi-class problems. comparison with well-known Bhattacharyya bound which is typically calculated by employing parametric assumptions, proposed paper are directly estimable from data, provably tighter, and more robust different types of data. We verify tightness validity using an illustrative synthetic example, further demonstrate its value incorporating it into a...

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

Purpose: The aim of this study was to leverage data-driven approaches, including a novel articulatory consonant distinctiveness space (ACDS) approach, better understand speech motor control in amyotrophic lateral sclerosis (ALS). Method: Electromagnetic articulography used record tongue and lip movement data during the production 10 consonants from healthy controls ( n = 15) individuals with ALS 47). To assess phoneme distinctness, were analyzed using two classification algorithms,...

Purpose: This study aimed to investigate the effect of stimulus signal length on tongue and lip motion pattern stability in speakers diagnosed with amyotrophic lateral sclerosis (ALS) compared healthy controls. Method: Electromagnetic articulography was used derive articulatory patterns from individuals mild ( n = 27) severe 16) ALS controls 25). The spatiotemporal index (STI) as a measure stability. Two experiments were conducted evaluate effects STI: (a) number syllables STI values (b)...

Purpose: The goal of this study was to examine the efficacy acceleration-based articulatory measures in characterizing decline speech motor control due amyotrophic lateral sclerosis (ALS). Method: Electromagnetic articulography used record tongue and lip movements during production 20 phrases. Data were collected from 50 individuals diagnosed with ALS. Articulatory kinematic variability measured using spatiotemporal index both instantaneous acceleration speed signals. Linear regression...

Estimating density functionals of analog sources is an important problem in statistical signal processing and information theory. Traditionally, estimating these quantities requires either making parametric assumptions about the underlying distributions or using non-parametric estimation followed by integration. In this paper we introduce a direct nonparametric approach which bypasses need for error rates k-NN classifiers as "data-driven" basis functions that can be combined to estimate...

Alan Wisler, Kristin Teplansky, Jordan Green, Yana Yunusova, Thomas Campbell, Daragh Heitzman, Jun Wang. Proceedings of the Eighth Workshop on Speech and Language Processing for Assistive Technologies. 2019.

This paper discusses the development of an active noise control (ANC) system to cancel compressor produced by a commercially available heating, ventilation and air conditioning unit enclosed within closet. Feedback ANC architecture that requires no reference microphone is used for cost-effectiveness. A novel delayless subband adaptive filtering technique reduce computational complexity algorithm improve performance. Finally, extended two-channel in order provide additional zone silence....

<b><i>Objective:</i></b> In the perceptual assessment of dysarthria, various approaches are used to examine accuracy listeners’ speech transcriptions and their subjective impressions disorder. However, less attention has been given effort cognitive resources required process samples. This study explores relationship between transcription accuracy, comprehensibility, speech, objective measures reaction time (RT) further challenges involved in processing dysarthric...

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease that affects bulbar functions including speech and voice. Voice onset time (VOT) was examined in speakers with ALS early late stages to explore the coordination of articulatory phonatory systems during production.VOT measured nonword /bap/ produced by early-stage (n = 11), late-stage 6), healthy controls 13), compared performance decline (a marker progression) ALS.Overall comparison VOT values among three groups showed...