Title :
Attractor dimension, entropy and modelling of speech time series
Author_Institution :
Dept. of Electr. Eng., Indian Inst. of Technol., Kanpur, India
Abstract :
Using the correlation dimension it is shown that most speech time series in the form of phoneme utterances have low-dimensional attractors. The positive values of second-order entropy show their chaotic nature. A reconstruction approach to the vocal tract and speech modelling problem is proposed.
Keywords :
entropy; speech analysis and processing; speech recognition; time series; correlation dimension; low-dimensional attractors; phoneme utterances; reconstruction approach; second-order entropy; speech modelling; speech time series; vocal tract;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19901147
