Computer-Implemented Articulatory Models for Speech Production: A Review.

Area of Science:

• Speech Production and Articulation Modeling
• Computational Linguistics
• Bioacoustics

Background:

• Speech production and articulation modeling is an evolving research area.
• Current articulatory models do not match the quality of corpus-based speech synthesis.
• Biomechanical and neuromuscular approaches are complex and not yet suitable for sentence-level synthesis.

Purpose of the Study:

• To review computer-implemented articulatory models and categorize them.
• To discuss neurobiological control mechanisms for articulatory models.
• To identify research needs for improving articulatory-acoustic models and speech synthesis.

Main Methods:

• Categorization of existing computer-implemented articulatory models.
• Detailed discussion of neurobiologically adequate control strategies for articulatory models.
• Analysis of current limitations in articulatory speech synthesis quality.

Main Results:

• A classification framework for articulatory models is presented.
• Key research questions regarding neural control of articulation are highlighted.
• The gap between current articulatory models and desired synthesis quality is identified.

Conclusions:

• Further development of articulatory-acoustic models is essential.
• These models will enable testing of neurobiological control concepts for speech.
• Advancements are needed to achieve high-quality, neurobiologically grounded speech synthesis.