Investigating the speech acquisition of hearing-impaired children attracts considerable attentions in recent years. Previous studies that investigate Mandarin-speaking children with hearing impairment mostly focus on production of some specific phonemes. Besides, the phonemes are sometimes embedded in a limited number of speech materials or uttered by only a few speakers. In this study, we analyzed the pronunciation errors of all Mandarin vowels and consonant produced by 60 pre- or post-lingually hearing-impaired children. We designed a set of speech materials that consisted of 153 monosyllable and 145 disyllable commonly used words and had a comprehensive phonetic coverage. The analysis shows that monophthongs were produced less accurately than diphthongs and triphthongs. Bilabials and nasals and plosives were relatively easier for hearing-impaired children to acquire than other consonants with respect to articulation manner and place, respectively. The Mandarin affricates had the lowest accuracy. Substitution is the most frequent error patterns for initial consonants while deletion is the common error for final consonants. The findings of this study can shed light on pronunciation teaching of hearing-impaired children. Besides, the corpus can benefit developing computer-assisted speech assessment system.
Cite as: Cheng, J., Yan, Y., Gao, Y., Feng, X., Wang, Y., Zhang, J. (2022) A study of production error analysis for Mandarin-speaking Children with Hearing Impairment. Proc. Interspeech 2022, 4840-4844, doi: 10.21437/Interspeech.2022-10477
@inproceedings{cheng22c_interspeech,
author={Jingwen Cheng and Yuchen Yan and Yingming Gao and Xiaoli Feng and Yannan Wang and Jinsong Zhang},
title={{A study of production error analysis for Mandarin-speaking Children with Hearing Impairment}},
year=2022,
booktitle={Proc. Interspeech 2022},
pages={4840--4844},
doi={10.21437/Interspeech.2022-10477}
}