Skip to main content
Weiter zur Homepage der Europäischen Kommission (öffnet in neuem Fenster)
Deutsch de
CORDIS - Forschungsergebnisse der EU
CORDIS
Inhalt archiviert am 2024-04-16

Articulatory-Acoustic Correlations in Coarticulatory Processes: a Cross-Language Investigation

Ziel

Recent theoretical and experimental work on speech production strongly indicates the need for a comprehensive cross-language programme of basic research to expand knowledge of the language-specific and language-independent regularities involved.
For many applications in speech technology the availability of an articulatory representation of speech utterances would mean substantial progress. For example, in the automatic recognition of continuous speech one of the major barriers to robust speaker-independent systems has been the great variability in the relationship between the acoustic level of representation and the phonological structure of a given utterance.
However, the variability can itself become a productive source of information if the underlying physiological and linguistic constraints of the dynamics of articulation can be modelled, and this is the goal of the ACCOR Action. The main source of systematic variability at the segmental level is undoubtedly coarticulation, the main focus of the work.
Modelling of the underlying physiological and linguistic constraints on the dynamics of articulation during speech production was undertaken. A cross-language study of coarticulation identified the major language-independent universal regularities of the phenomenon and how they interact with language-specific factors. Investigation of coarticulatory regularities were integrated with new and improved ways of exploiting these regularities in deriving articulatory representations from the acoustic analysis of speech. Hardware and software were developed for digital recording of articulatory and acoustic data. A PC-based prototype of a multichannel speech workstation has been designed. Research was conducted into the instrumental investigation of articulatory activity. A review was made of connected speech processes in English, French, German and Italian. A core lexicon for English, French, German, Italian, Gaelic Irish, Catalan and Swedish has been established, consisting of consonant-vowel-consonant nonsense words, real words matching the phonetic structure of the nonsense words as closely as possible, and sentences illustrating the main connected speech processes in different languages. A complete set of articulatory, aerodynamic and acoustic recordings is available involving 5 speakers of each language. A common methodology has been adopted for segmenting and labelling the data. It is based on a nonlinear and independent annotation of the articulatory and acoustic events. Articulatory and acoustic data were interpreted in terms of an integrated production model.
APPROACH AND METHODS
A cross-language approach was adopted in the ACCOR Action as a means of identifying the major language-independent universal regularities of the phenomenon (aspects such as mass, inertia and elasticity of the speech organs, the mechanical linkages betweenthem and the neuromuscular complexities of the cranial nervous system), and how they interact with language-specific factors such as the phonological rules of the languages concerned.
Work began by examining the activities of the major physiological systems underlying speech production: the respiratory system (producing a flow of air), the laryngeal system (modifying the airflow by the valving mechanism of the vocal folds), and the complex supraglottal structures in the mouth and nose (such as the tongue, lips, jaw and soft palate, which shape the vocal tract into different resonating cavities).
In addition, by examining the details of a given articulatory process for a number of different languages, ACCOR aimed to determine how such processes differ according to different phonological systems, and thus to be in a position to investigate interactions between the two sources of variation. A common methodology, standardised investigation tools and measurement procedures at specified locations in the speech signal have been adopted.
PROGRESS AND RESULTS - STATUS OF OCTOBER 1991
Deliverables and papers cover the following areas:
-Hardware and software development for digital recording of articulatory and acoustic data. A PC-based prototype of a multichannel speech workstation has been designed. This equipment will be commercially available in early 1992.
-Research into the instrumental investigation of articulatory activity.
-Review of connected speech processes in English, French, German and Italian.
-A core lexicon for English, French, German, Italian, Gaelic Irish, Catalan and Swedish has been established, consisting of consonant-vowel-consonant nonsense words; real words matching the phonetic structure of the nonsense words as closely as possible; and sentences illustrating the main connected speech processes in different languages. A complete set of articulatory, aerodynamic and acoustic recordings is available involving 5 speakers of each language.
-A common methodology has been adopted for segmenting and labelling the data. It is based on a non-linear and independent annotation of the articulatory and acoustic events.
-The interpretation of articulatory and acoustic data in terms of an integrated production model.
POTENTIAL
This Action integrates investigations into coarticulatory regularities with research into new and improved ways of exploiting these regularities in deriving articulatory representations through the acoustic analysis of speech. The knowledge gained will show the most promising paths to follow in many speech technology applications, particularly in the development of more robust speech recognition systems.

Wissenschaftliches Gebiet (EuroSciVoc)

CORDIS klassifiziert Projekte mit EuroSciVoc, einer mehrsprachigen Taxonomie der Wissenschaftsbereiche, durch einen halbautomatischen Prozess, der auf Verfahren der Verarbeitung natürlicher Sprache beruht. Siehe: Das European Science Vocabulary.

Sie müssen sich anmelden oder registrieren, um diese Funktion zu nutzen

Programm/Programme

Mehrjährige Finanzierungsprogramme, in denen die Prioritäten der EU für Forschung und Innovation festgelegt sind.

Thema/Themen

Aufforderungen zur Einreichung von Vorschlägen sind nach Themen gegliedert. Ein Thema definiert einen bestimmten Bereich oder ein Gebiet, zu dem Vorschläge eingereicht werden können. Die Beschreibung eines Themas umfasst seinen spezifischen Umfang und die erwarteten Auswirkungen des finanzierten Projekts.

Daten nicht verfügbar

Aufforderung zur Vorschlagseinreichung

Verfahren zur Aufforderung zur Einreichung von Projektvorschlägen mit dem Ziel, eine EU-Finanzierung zu erhalten.

Daten nicht verfügbar

Finanzierungsplan

Finanzierungsregelung (oder „Art der Maßnahme“) innerhalb eines Programms mit gemeinsamen Merkmalen. Sieht folgendes vor: den Umfang der finanzierten Maßnahmen, den Erstattungssatz, spezifische Bewertungskriterien für die Finanzierung und die Verwendung vereinfachter Kostenformen wie Pauschalbeträge.

Daten nicht verfügbar

Koordinator

Centre National de la Recherche Scientifique (CNRS)
EU-Beitrag
Keine Daten
Adresse
1919 route de Mende
34405 Montpellier
Frankreich

Auf der Karte ansehen

Gesamtkosten

Die Gesamtkosten, die dieser Organisation durch die Beteiligung am Projekt entstanden sind, einschließlich der direkten und indirekten Kosten. Dieser Betrag ist Teil des Gesamtbudgets des Projekts.

Keine Daten

Beteiligte (8)

Mein Booklet 0 0