Codebook Constrained Wiener Filtering for Speech Enhancement - Speech and Audio Processing, IEEE Transactions on

We propose a modified hidden Markov model (MHMM) thatincorporates nonparametric state duration and state duration-dependentobservation probabilities to reflect state transitions and to have accuratetemporal structures in the HMM.In addition, to cope with the problem that results from the use ofinsufficient amount of training data, we propose to use the modifiedcontinuous density hidden Markov model (MCDHMM) with a differentnumber of mixtures for the probabilities of state duration-independentand state duration-dependent observation. We show that this proposedmethod yields improvement in recognition accuracy in comparison withthe conventional CDHMM.