Speaker invariant principal component determination for Bangla vowel perceptual space

2016 5th International Conference on Informatics, Electronics and Vision (ICIEV)(2016)

引用 0|浏览1
暂无评分
摘要
In this article, speaker invariant principal components have been determined from the multi-dimensional perceptual space for Bangla vowels. Linguistic content base nine-dimensional perceptual space has been formed from the formant frequency related properties of ten Bangla vowels (/a/, /ã/, /i/, /ĩ/,/u/, /ũ/ /e/, /ai/,/õ/, and /au/). Average normalized eigen-value profile is calculated using Bangla vowel perceptual spaces of fifteen speakers. Speaker invariance is numerically evaluated using the correlation of first and second principal components of the fifteen speakers. The contributions of perceptual vector members in first and second principal components are also investigated.
更多
查看译文
关键词
Bangla vowel perceptual spaces,Speaker invariant,Perceptual vector members,Linguistic content,Eigen-value and eigenvector,Vocal-tract,Correlation
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