Title :
Estimation of the dynamics of vocal tract parameters
Author_Institution :
Siemens AG, München, Germany
Abstract :
The dynamics of the vocal tract are modelled by a perturbation, which changes randomly the area function of the vocal tract. Wakita´s method establishes a correspondence between the perturbation of the area function and a perturbation of the prediction coefficients. A Kalman filter extracts continuously the estimates of the prediction coefficients using the structure of the perturbation of the area coefficients.
Keywords :
Acoustic propagation; Boundary conditions; Covariance matrix; Elasticity; Filters; Linear predictive coding; Shape; Speech analysis; Speech synthesis; State estimation;
Conference_Titel :
Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '81.
DOI :
10.1109/ICASSP.1981.1171267