Objective Evaluation Of Glottal Stops In Cleft Palate Speech Based On Spectral Moments

INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND SOFTWARE ENGINEERING (AISE 2014)(2014)

引用 0|浏览2
暂无评分
摘要
This study investigated acoustic features of glottal stops produced as substitutes for stop consonants by Mandarin speakers with cleft palate. As indices for overall energy distribution of the spectra, spectral moments were derived for the burst and the subsequent portion of stop consonants. The results show that stop consonants that were produced as glottal stops had less acoustic energy in the high-frequency range than typical stop productions. This finding might be due to the fact that the anterior cavity for glottal stops is usually larger than the one for typical stops. We also found that the spectral energy of glottal stops spread over a wider frequency range. Logistic regression was proposed as a pattern classifier and we obtained acceptable classification accuracy when selecting an optimal set of a few spectral coefficients as input factors. It appears that it is possible to objectively assess the presence of glottal stops with such spectral indices and classifier, and the assessment can be further automated if combining an automatic detection of stop bursts.
更多
查看译文
关键词
Cleft palate speech, Glottal stop, Spectral moment, Logistic regression
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要