La modelización de los elementos suprasegmentales para la síntesis

Esta página no se actualiza con regularidad



La modelización de los elementos suprasegmentales para la síntesis

La síntesis de los elementos suprasegmentales

Los sistemas de conversión de texto a habla incluyen una serie de módulos encargados de la síntesis de los elementos suprasegmentales:

Los módulos entonativos:

Definición de “modelo melódico”:

Elaboración de modelos entonativos

Aproximaciones fonéticas y fonológicas:

Tal como Cutler y Ladd (1983) señalan, existen dos tipos de aproximaciones al estudio de la entonación:

“There are two broad traditions in the study of prosody that may be characterized - or caricatured - by their methodological preferences for one or the other of the scientific activities mentioned in the title: making measurements and constructing models. On one side of the dichotomy stand instrumental and experimental studies that seek to quantify acoustic features and investigate perceptual responses. On the other are descriptive and theoretical studies of prosodic structure and its relation to other aspects of grammar and phonology. In a great deal of past work these two traditions have simply ignored one another” (Cutler y Ladd, 1983, p. 1).

(a) Aproximación fonética: estudia la entonación desde el punto de vista de su manifestación física.

“for many of those who take the measurer's approach, the primary concern is not representation, but realization. The question being asked is: What are the physical correlates of this or that prosodic message? To the extent that such investigators have constructed explicit models of prosodic representation, they have tended to think in terms not of linguistic categories, but of interacting parameters; their models assign acoustic correlates to individual functions such as word stress, sentence stress, sentence modality, affective use of pitch range, and so on, and attempt to specify the interaction of all these effects on individual parameters like fundamental frequency” (Cutler y Ladd, 1983, p. 5).

(b) Aproximación fonológica: se interesa más por las estructuras lingüísticas que subyacen tras esas manifestaciones físicas:

“The model-builder is interested in establishing an inventory of abstract categories - a formal representation - of prosodic function and prosodic form. The goal of the model-builder's enterprise is to describe the systematic structure underlying prosodic distinctions; the basic assumption is that there are well-defined abstract levels of representation that mediate between specific prosodic functions like p“hrase boundary” and specific acoustic traits” (Cutler y Ladd, 1983, p. 5).

Independientemente del tipo de aproximación, el análisis y modelización de la melodía implica una serie de etapas y tareas:

etapas.GIF

Representación esquemática de las diferentes etapas del análisis experimental (Garrido, 1996)


La modelización de los elementos suprasegmentales para la síntesis
Juan M. Garrido, Universitat Autònoma de Barcelona

Last updated: 21 August 2003
Esta página no se actualiza con regularidad