A study of non-native accent correction techniques combining phonetics, machine learning and biomechanics

Yanziye WEI*

*Corresponding author for this work

Research output: Journal PublicationsJournal Article (refereed)peer-review

Abstract

This study provides an in-depth discussion of non-native accent correction techniques, combining phonological principles with insights from biomechanics and machine learning algorithms. By examining the physical aspects of speech production, such as articulatory movements and vocal tract dynamics, the research highlights how biomechanical factors influence the pronunciation characteristics of non-native speakers. The study reports on the current state of the art in accent correction technology, detailing how biomechanical analysis can enhance the understanding of speech patterns and contribute to more effective correction techniques. Experimental investigations verify the effectiveness of these methods across different language contexts, demonstrating significant improvements in pronunciation accuracy, fluency, and user satisfaction. By incorporating biomechanical principles, this research provides a new theoretical basis and technical support for the field of non-native accent correction, which is of positive significance for the promotion of cross-cultural communication, as they address the physical challenges faced by non-native speakers in articulating sounds specific to different languages.
Original languageEnglish
Article number725
JournalMCB Molecular and Cellular Biomechanics
Volume22
Issue number1
Early online date10 Jan 2025
DOIs
Publication statusPublished - 10 Jan 2025

Bibliographical note

Publisher Copyright:
© 2025 by author(s).

Keywords

  • biomechanics
  • correction techniques
  • feature extraction
  • machine learning
  • netics
  • non-native accents

Fingerprint

Dive into the research topics of 'A study of non-native accent correction techniques combining phonetics, machine learning and biomechanics'. Together they form a unique fingerprint.

Cite this