This study investigated the effectiveness of utilizing acoustic measures as an objective tool in monitoring speech errors and providing visual feedback to enhance speech training and aural rehabilitation of children with hearing impairment. The first part of the study included a comprehensive description of the acoustic characteristics related to the speech deficits of a hearing impaired child. Results of a series of t-tests performed on the experimental measures showed that vowel length and the loci of formant frequencies were most relevant in differentiating between correctly and incorrectly produced vowels, while voice onset time along with measures of Moment 1 (mean) and Moment 3 (skewness) obtained from speech moment analysis, were related to consonant accuracy. These findings, especially the finding of an abnormal sound frequency distribution shown in the hearing impaired child's consonant production, suggest a link between perceptual deficits and speech production errors and provide clues to the type of compensatory feedback needed for aural rehabilitation. The second part of the study involved a multiple baseline design across behaviours with replication across three hearing impaired children to assess the efficacy of treatment with acoustic signals as visual feedback. Participants' speech articulations following traditional speech training and training using spectrographic and RMS displays as visual feedback (referred to as 'visual treatment') were compared, with traditional non-visual treatment followed by visual treatment on one or two targets in a time-staggered fashion. Although no statistically significant difference on the experimental measures was found between the two training approaches based on perceptual assessment, some objective acoustic measures revealed more subtle changes toward normal speech patterns with visual treatment as compared to a traditional approach. Further acoustic-perceptual studies with a larger sample size and longer experimental period are needed to better understand the general and long-term effectiveness of visual treatment.

This study investigated the effectiveness of utilizing acoustic measures as an objective tool in monitoring speech errors and providing visual feedback to enhance speech training and aural rehabilitation of children with hearing impairment. The first part of the study included a comprehensive description of the acoustic characteristics related to the speech deficits of a hearing impaired child. Results of a series of t-tests performed on the experimental measures showed that vowel length and the loci of formant frequencies were most relevant in differentiating between correctly and incorrectly produced vowels, while voice onset time along with measures of Moment 1 (mean) and Moment 3 (skewness) obtained from speech moment analysis, were related to consonant accuracy. These findings, especially the finding of an abnormal sound frequency distribution shown in the hearing impaired child's consonant production, suggest a link between perceptual deficits and speech production errors and provide clues to the type of compensatory feedback needed for aural rehabilitation. The second part of the study involved a multiple baseline design across behaviours with replication across three hearing impaired children to assess the efficacy of treatment with acoustic signals as visual feedback. Participants' speech articulations following traditional speech training and training using spectrographic and RMS displays as visual feedback (referred to as 'visual treatment') were compared, with traditional non-visual treatment followed by visual treatment on one or two targets in a time-staggered fashion. Although no statistically significant difference on the experimental measures was found between the two training approaches based on perceptual assessment, some objective acoustic measures revealed more subtle changes toward normal speech patterns with visual treatment as compared to a traditional approach. Further acoustic-perceptual studies with a larger sample size and longer experimental period are needed to better understand the general and long-term effectiveness of visual treatment.

The purpose of this study was to examine the effect of fundamental frequency (Fo) on stop consonant voice onset time (VOT). VOT was measured from the recordings of 56 young men reading phrases containing all 6 English voiced and voiceless stops in word-initial position across high-, medium-, and low-Fo levels. Separate analyses of variance for the voiced and voiceless stops revealed no significant main effect for Fo for the voiced stops but a significant Fo effect for the voiceless stops. Across the voiceless stops, productions at high Fos displayed significantly shorter VOTs than productions at low or mid F os. The findings indicated that researchers must take into account the Fo level at which voiceless stop VOT is measured.

No description available.

Much of speech kinematics research is conducted by attaching a device to the articulators. However very little research has been conducted to determine what influence these devices may have on the perceptual and acoustic characteristics of speech. This study examined the effect of placing a small magnet on the tongue of ten normal adult speakers while reading a sentence containing /s/ and "sh" in initial, medial and final position. Two different placements of 10 and 15 mm from the tip of the tongue were analyzed. Data were taken before magnet placement, immediately after magnet placement, after 5 minutes of conversation, and after an additional 10 minutes of conversation. The acoustic output was analyzed using spectral moments analysis (spectral mean, variance, skewness, and kurtosis). Changes in spectral mean and variance were found for "sh" as a result of magnet placement, which was characterized by an interaction effect between condition and the word position of the target fricative. In addition, significant changes in spectral mean were found for /s/ and "sh" as a result of magnet position. Although results from the present study indicated that there were some acoustic changes in fricative productions with a marker attached at midline, the spectral changes were not consistent or pervasive, and speakers were able to adapt to the presence of the magnet in a relatively short amount of time.

Background: Technological advancements in speech acoustic analysis have led to the development of spectral/cepstral analyses due to questions regarding the validity of traditional time-based measures (i.e., Jitter, Shimmer, and Harmonics-to-Noise-Ratio) in objectifying perturbations in dysphonic voices. Aim: This study investigated the validity of time-based measures in discriminating those with Friedreich’s ataxia (FA) from normal voiced (NV) peers when compared to cepstral-spectral measures. Method: A total of 120 sustained vowel phonations from an existing database of 40 participants (20 FA; 20 NV) of the vowels /ɑ/, /i/, and /o/ were analyzed to determine which set of variables (i.e., time-based vs. cepstral-spectral) better predicted group membership. Four variables of time-based measures (Jitter Local %, Jitter RAP, Shimmer Local %, Shimmer APQ11, and HNR) were analyzed via the freeware program PRAAT and compared to four cepstral-spectral measures (Cepstral Peak Prominence, Cepstral Peak Prominence Standard Deviation, Low/High Ratio Standard Deviation, and the Cepstral/ Spectral Index of Dysphonia) extracted from the Analysis of Dysphonia in Speech and Voice (ADSV) software program. Results: Findings from a discriminant analysis showed sensitivity and specificity results to be better for ADSV measures; 100% of those in the FA group were classified correctly (sensitivity), and 95% of members in the NV group were correctly identified (specificity) as compared to PRAAT (70% sensitivity and 85% specificity). Conclusions: Cepstral-spectral measures are much more accurate in discriminating between those with FA and NV peers as compared to time-based estimates.

Les opérateurs de télécommunications ont besoin de superviser en temps réel la qualité vocale des services qu'ils proposent. La qualité vocale peut être évaluée par tests subjectifs auprès d'utilisateurs; mais ces méthodes sont très coûteuses et peu adaptées à la supervision. Des modèles objectifs sont ainsi proposés afin de prédire la qualité vocale à moindre coût. Cette thèse propose un modèle de diagnostic et d’évaluation utilisant les informations disponibles au point de mesure : le modèle DESQHI (Diagnostic and Speech Quality using Hybrid Indicators). Il se distingue des modèles existants par deux caractéristiques principales. La première concerne la structure du cœur du modèle. Il est montré que la qualité vocale peut être représentée comme un phénomène multidimensionnel faisant intervenir trois dimensions perceptives correspondant à bruyance, codage de la parole et continuité. Cette structure permet de diagnostiquer la qualité vocale en identifiant les principales causes perceptives de sa dégradation. La deuxième caractéristique concerne le type d’indicateur utilisé pour représenter ces dimensions perceptives, à savoir l’utilisation d’indicateurs basés sur le signal et paramétriques. Les indicateurs basés sur le signal utilisent les informations numériques pour représenter les caractéristiques du signal (par exemple le rapport signal sur bruit qui donne une estimation du niveau sonore du bruit de fond). Les indicateurs paramétriques sont issus des statistiques du réseau (par exemple le pourcentage de pertes de paquets qui fournit une indication sur le niveau de discontinuité du signal de parole). L’utilisation d’indicateurs hybrides utilisant à la fois les informations du signal numérique et les statistiques du réseau permet d’améliorer les performances globales de la prédiction de la qualité vocale, comparativement aux modèles uniquement basés sur le signal (p. ex. modèle P.563) et aux modèles utilisant les indicateurs paramétriques (p. ex. modèle E). / With increasing development of new technologies (RTC, RNIS, GSM, VoIP), tele-communication services are becoming more and more diversified. To this end, telecommunication operators need to supervise in real-time the speech quality of the services they offer. Speech quality is usually evaluated from subjective experiments.. Nevertheless, such experiments are time consuming and do not allow any supervisory control. So, accurate objective models are useful to estimate the speech quality.This thesis proposes a non-intrusive model for diagnosing and evaluating speech quality using information available at the measurement point: the DESQHI model (Diagnostic and Evaluation of Speech Quality using Hybrid Indicators). It differs from existing models in terms in two main characteristics. The first one concerns the structure of the model. It is shown that speech quality can be represented as a multidimensional phenomenon incorporating three perceptual dimensions related to noisiness, speech codec and continuity. This multidimensional structure allows for a diagnostic of speech quality based on identifying the principal features affecting speech qual-ity. The second characteristic concerns the nature of indicators (signal-based and parametric) used to represent the three perceptual dimensions. Signal-based indicators use numeric information to represent the characteristics of the signal, for example, the loudness of the speech signal. Parametric indicators are obtained from the network statistics, for example, the percentage of packet loss, which gives information about the level of the discontinuity in the speech signal. This work proposes hybrid indicators (using both signal-based and parametric metrics). It is shown that they are better speech quality predictors than existing models, either parametric only (e.g. ITU-T Recommendation G.107, also known as the E-model) or signal-based only (e.g. ITU-T Recommendation P.563 model).

Une limitation fondamentale du problème inverse acoustique est déterminée par la taille et la densité de l'antenne de microphones. Une solution pour atteindre une grande antenne et / ou à forte densité de microphones est de scanner l'objet d'intérêt en déplaçant séquentiellement une antenne de petites dimensions (mesures dites séquentielles). La différence entre des mesures séquentielle et une mesure simultanée en tous points est que dans ce dernier cas, l'intégralité de la matrice interpectrale peut être estimée, contrairement au cas des mesures séquentielles qui ne permettent l'estimation que d'une partie réduite de cette matrice; les éléments croisés (interspectres) entre deux points de mesures n'appartenant pas à la même séquence ne sont pas estimés. Néanmoins, ces données restent nécessaires pour la reconstruction acoustique. Dans l'approche classique, une ou plusieurs références sont utilisées pour retrouver les données manquantes. L'objet de cette thèse est de récupérer les éléments manquants de la matrice interspectrale sans capteurs de référence, dans le cas où le champ acoustique est suffisamment cohérent pour mettre en œuvre les mesures séquentielles. Deux modèles de spectre de valeurs propres parcimonieux sont proposés pour résoudre ce problème, le premier impose une valeur faible de rang, tandis que le second repose sur la minimisation de la norme nucléaire (recherche d'une solution faiblement parcimonieuse). / A fundamental limitation of the inverse acoustic problem is determined by the size of the array and the microphone density. A solution to achieve large array and/or high microphone density is to scan the object of interest by moving sequentially a small prototype array, which is referred to as sequential measurements. In comparison to a large array and/or high microphone density array that can acquire simultaneously all the information of the spectral matrix, in particular all cross-spectra, sequential measurements can only acquire a block diagonal spectral matrix, while the cross-spectra between the sequential measurements remain unknown due to the missing phase relationships between consecutive positions. Nevertheless, these unknown cross-spectra are necessary for acoustic reconstruction. The object of this thesis is to recover the missing elements of the spectral matrix in the case that the acoustical field is highly coherent so as to implement the sequential measurements. Sparse eigenvalue spectrum are assumed to solve this problem, which lead to a structured low rank model and a weakly sparse eigenvalue spectrum model.

Dans un environnement naturel, le champ acoustique est complexe (plusieurs sources, différentes positions spatiales, acoustique du lieu,...) et l'écoute est binaurale. Le filtrage acoustique opéré par la tête, le buste et les pavillons de l'auditeur (dépendant de la direction) induit donc systématiquement des différences interaurales de temps, de niveau et de spectre. Des modèles de sonie existent et permettent de prédire la sonie des sons stationnaires dans des situations d'écoute simples (ISO-532B (1975), DIN-45631 (1990), ANSI-S3.4 (2007)). L'écoute doit être monaurale (une seule oreille) ou diotique (même son aux deux oreilles), correspondant à une source en incidence frontale en champ libre, ou en champ diffus. En revanche, ces modèles échouent pour prédire la sonie lorsque des différences interaurales importantes interviennent. La thèse s’est ainsi intéressée à la sonie des champs acoustiques stationnaires, impliquant une ou plusieurs sources, soit artificielles et spatialisées en champ libre, soit réelles dans une acoustique naturelle. De nouveaux éléments ont été apportés dans la compréhension dont l'information contenue dans les signaux reçus aux oreilles de l’auditeur est combinée pour former un unique percept de sonie binaurale dans les situations d’écoute dichotiques (gain de sommation binaural, cas de plusieurs sources, effet de la corrélation interaurale,…). D’une part, une hypothèse pour essayer d’expliquer les différences interindividuelles observées dans les stratégies de sommation binaurales a pu être testée ; la robustesse et la stabilité au cours du temps de ces stratégies individuelles a été mise en avant. D’autre part, trois principaux modèles psychophysiques de sonie binaurale (ANSI-S3.4 (2007), Moore et Glasberg (2007), Sivonen et Ellermeier (2008)) ont été testés sur l’ensemble des données expérimentales (impliquant différents niveaux de réalisme), permettant de préciser la performance et les domaines de validité respectifs de chacun de ces modèles dans des situations d’écoute fortement dichotiques. / Listening in a natural environment implies to consider complex sound fields (several sound sources, different spatial positions, reflections…) and a binaural listening configuration. As a consequence, differences in time, level and spectrum between the two at-ear signals are systematically induced by the direction-dependant physical filtering from the human head, torso and pinnae. Existing loudness models provide accurate predictions under simplified listening situations (ISO-532B (1975), DIN-45631 (1990), ANSI-S3.4 (2007)). These models have been designed to use monaural (only one ear) or diotic (same signal at the two ears) signals, equivalent to one sound source with a frontal incidence, in a free or diffuse field. However, the models fail to predict loudness when the interaural differences are large. The present document focuses on the loudness of stationary sound fields, made up of one or several, artificial or real sound sources, in a free field or in a real environment. New elements have been brought to light regarding how, in a dichotic listening situation, information is combined from the two ears to produce one unique binaural loudness percept (binaural gain, case of several sound sources, effect of interaural correlation,...). On the one hand, one hypothesis have been tested in order to try to explain the interindividual differences observed in binaural loudness summation ; the robustness and stability over time of these individual strategies have been highlighted. On the other hand, predictions from the three main psychophysical models of binaural loudness (ANSI-S3.4 (2007), Moore et Glasberg (2007), Sivonen et Ellermeier (2008)) have been compared with all of the subjective data (involving different levels of realism), which allowed us to define more accurately the domain of validity and performance of these models in highly dichotic listening situations.

Dans le cadre de la lutte en mer, la détection acoustique des structures immergées s’effectue généralement sur des signaux stationnaires. Une nouvelle génération de sonars permet de détecter sur des signaux transitoires. Ceci implique de compléter le processus de conception des projets industriels qui ne tient compte d’exigences qu'en matière de bruits rayonnés en régime stationnaire. Il est donc nécessaire de comprendre les mécanismes de transfert des sources de bruit transitoires sur les structures immergées. Cette thèse s’inscrit dans ce cadre et consiste à étudier les mécanismes de transfert vibratoire et de rayonnement acoustique qui peuvent intervenir sur ces structures lorsque l’excitation est transitoire. L’analyse porte sur différents éléments de la chaine de transfert : le rayonnement dans l’eau du bordé, la diffraction des ondes par les raidisseurs, et le comportement résonnant des structures internes supportant les matériels. Le premier chapitre présente une analyse bibliographique autour de l’étude des phénomènes vibroacoustiques transitoires des structures immergées, de l’influence d’un fluide lourd sur le comportement vibroacoustique des plaques, et des méthodes de calcul vibroacoustiques en régime transitoire. Dans le second chapitre nous étudions la réponse transitoire d’une plaque infinie immergée soumise à une force impulsionnelle ponctuelle. La méthode de calcul s’appuie sur les calculs spectraux fréquences-nombre d’onde. Les réponses temporelles sont obtenues par transformées de Fourier inverses. L’analyse des spectres et des réponses temporelles de l’accélération vibratoire de la plaque et de la pression rayonnée, met en évidence l’influence de la présence du fluide. La prise en compte de l’inertie rotationnelle et du cisaillement à travers le modèle de plaque de Mindlin-Timoshenko est également étudiée. Ces résultats sont confrontés à une expérimentation présentée dans le troisième chapitre. La structure étudiée est une plaque rectangulaire posée horizontalement à la surface d’une cuve remplie d’eau. Deux types de sources transitoires sont utilisés : marteau de choc, lâché d’une bille. La comparaison des résultats numériques et expérimentaux montre que l’on retrouve certains phénomènes évoqués précédemment. L’effet des raidisseurs sur le rayonnement acoustique fait l’objet du quatrième chapitre. Une plaque raidie périodiquement dans une direction est considéré. L’influence des ondes de Bloch-Floquet sur la réponse temporelle est étudiée. Les résultats sont comparés à des mesures effectuées sur une barge d’essais. Dans le cinquième chapitre, l’effet des structures internes est étudié à partir d’un modèle de plaque couplé à un système résonnant constitué d’un assemblage poutre-plaque. La méthode des inertances est utilisée pour obtenir les forces de couplage entre les différents éléments. Les signaux temporels sont étudiés en fonction de l’importance de la rupture d’inertance entre la plaque et l’assemblage. / In the sea, the acoustic detection of other battle engines is done by detecting mostly stationary signals. However, new types of detection systems are being developed, and are able to detect and to analyze transient signals. Therefore, the industrial conception process needs to be improved, so that the underwater vehicles transient noises can be taken in account. In order to do so, the mechanism of vibroacoustic transfer of transient sources of submerged structures has to be understood. The object of this thesis is then the study of the mechanism of vibration transfer and acoustic radiation of those structures when they are excited by transient sources. The shell radiation in the water, the wave diffraction by circumferential stiffeners and the resonant behavior of internal substructures are analyzed. The first chapter presents the bibliographical study of three themes: the study of transient phenomenon of submerged structures, the influence of heavy fluid coupling on vibroacoustic behavior of plates, and the different calculation methods in transient vibroacoustics. In the second chapter, we study the transient response of a submerged infinite plate excited by an impulsively point force. First, the calculations are done in the wavenumber-frequency domain. Then the spatio-temporal responses are obtained using inverse Fourier transforms. The discretization of wavenumber and frequency domains and the damping model are studied. The analysis of frequency and time responses of the plate vibration and the radiated pressure enable us to observe the influence of heavy fluid coupling. Besides the Mindlin-Timoshenko plate model is also used and the effect of rotation inertia and shear stress are studied. In the third chapter, these numerical results are confronted to experimental data, obtained experimentally. The studied structure is a rectangular plate lying on the surface of a water tank. Two different excitations are used: an impact hammer and the free fall of a steel ball. The study of the correlation between numerical and experimental results showed that some phenomena are observed in both cases. The influence of stiffeners on the acoustic radiation is the theme of the fourth chapter. An infinite plate which is periodically stiffened through one direction is considered. The effect of Bloch-Floquet waves on time response is studied. Numerical results are compared to measurements data obtained on an industrial submerged structure. In the fifth chapter, the effect of internal structures is analyzed by modelling an infinite plate coupled to a resonant system made of a beam and a rectangular finite plate. The inertance coupling method is used to obtain the coupling forces between the different substructures. Influence of inertance difference between the substructures is illustrated by the time signals.