Date Available
7-27-2020
Year of Publication
2020
Degree Name
Master of Science in Electrical Engineering (MSEE)
Document Type
Master's Thesis
College
Engineering
Department/School/Program
Electrical and Computer Engineering
First Advisor
Dr. Michael T. Johnson
Abstract
This work presents the development, implementation, and evaluation of a Mispronunciation Detection and Diagnosis (MDD) system, with application to pronunciation evaluation of Mandarin-accented English speech. A comprehensive detection and diagnosis of errors in the Electromagnetic Articulography corpus of Mandarin-Accented English (EMA-MAE) was performed by using the expert phonetic transcripts and an Automatic Speech Recognition (ASR) system. Articulatory features derived from the parallel kinematic data available in the EMA-MAE corpus were used to identify the most significant articulatory error patterns seen in L2 speakers during common mispronunciations. Using both acoustic and articulatory information, an ASR based Mispronunciation Detection and Diagnosis (MDD) system was built and evaluated across different feature combinations and Deep Neural Network (DNN) architectures. The MDD system captured mispronunciation errors with a detection accuracy of 82.4%, a diagnostic accuracy of 75.8% and a false rejection rate of 17.2%. The results demonstrate the advantage of using articulatory features in revealing the significant contributors of mispronunciation as well as improving the performance of MDD systems.
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2020.340
Recommended Citation
Khanal, Subash, "MISPRONUNCIATION DETECTION AND DIAGNOSIS IN MANDARIN ACCENTED ENGLISH SPEECH" (2020). Theses and Dissertations--Electrical and Computer Engineering. 156.
https://uknowledge.uky.edu/ece_etds/156
