Objective Measures of Plosive Nasalization in Hypernasal Speech

Abstract

Hypernasal speech is a common symptom across several neurological disorders; however it has a variable acoustic signature, making it difficult to quantify acoustically or perceptually. In this paper, we propose the nasal cognate distinctiveness features as an objective proxy for hypernasal speech. Our method is motivated by the observation that incomplete velopharyngeal closure changes the acoustics of the resultant speech such that alveolar stops /t/ and /d/ map to the alveolar nasal /n/ and bilabial stops /b/ and /p/ map to bilabial nasal /m/. We propose a new family of features based on likelihood ratios between the plosives and their respective nasal cognates. These features are based on an acoustic model that is trained only on healthy speech, and evaluated on a set of 75 speakers diagnosed with different dysarthria subtypes and exhibiting varying levels of hypernasality. Our results show that the family of features compares favorably with the clinical perception of speech-language pathologists subjectively evaluating hypernasality.

Say What? A Dataset for Exploring the Error Patterns That Two ASR Engines Make

“…Goodness of Pronunciation [9] is a measure of articulatory precision that compares the log probability of a "true" phoneme from a transcript to the most likely phoneme given a frame of speech data, as assessed by the internal probabilities of an ASR system. This and related metrics have been applied in both educational [10] and clinical contexts [11].…”

Robust Estimation of Hypernasality in Dysarthria with Acoustic Model Likelihood Features

“…Features based on automatic speech recognition (ASR) acoustic models targeting articulatory precision have been used in nasality assessment systems [48]. The nasal cognate distinctiveness measure uses a similar approach but assesses the degree to which specific stops sound like their co-located nasal sonorants [56].…”

“…As expected, there is a decrease in the articulation precision feature value for an increase in severity of hypernasality. Furthermore, we expect hypernasality to exhibit unique patterns in terms of affected and unaffected unvoiced phonemes, that are not general to dysarthria [56], making phoneme-level AP classification a valuable signal in quantifying hypernasality.…”

Planting the seed for sound change: Evidence from real-time MRI of velum kinematics in German

“…Understanding how such production-perception relationships could give rise to sound change requires detailed knowledge of the different ways in which a vowel and a following nasal consonant variably overlap in speech production, especially with regard to the kinematics of the velum gesture associated with the nasal consonant. Despite some progress in identifying nasalization from the acoustic signal using measures such as the relative amplitudes of nasal and oral formants (A1-P0 and A1-P1; Beddor 2009, Chen 1997, Zellou 2017, accurately tracking the time course and magnitude of velum raising and lowering from speech acoustics both across speakers and within individual tokens is notoriously difficult (Barlaz et al 2018, Carignan 2018b, Feng & Castelli 1996, Saxon et al 2019, Styler 2017. Moreover, acoustic measures of this type are useful only for examining nasalization in vowels, where oral formants are present alongside nasal formants in the acoustic spectrum; such measures therefore cannot be used to identify and characterize nasalization throughout an entire ṼN(C) sequence.…”