This paper presents a new parameter estimation algorithm

based on the Extended Kalman Filter (EKF) for the recently

proposed statistical coarticulatory Hidden Dynamic

Model (HDM).We show how the EKF parameter estimation

algorithm unifies and simplifies the estimation of both the

state and parameter vectors. Experiments based on N-best

rescoring demonstrate superior performance of the (contextindependent)

HDM over a triphone baseline HMM in the

TIMIT phonetic recognition task. We also show that the

HDM is capable of generating speech vectors close to those

from the corresponding real data.

PDF file

In  Proc. of the Int. Conf. on Acoustics, Speech, and Signal Processing

Details