Parameter Estimation for a Maximum-entropy Acoustic Model
نویسنده
چکیده
The author has recently proposed maximum-entropy hidden Markov models (MEHMMs) as an acoustic model for speech recognition [3]. Maximum-likelihood parameter estimation for MEHMMs poses a signi cant computational challenge, and so the experiments in that paper were arti cially limited by requiring • that the training data include the hidden state 1 sequence, and • that there be only a limited number of permissible state sequences (a few hundred) per utterance.
منابع مشابه
CREDIBILISTIC PARAMETER ESTIMATION AND ITS APPLICATION IN FUZZY PORTFOLIO SELECTION
In this paper, a maximum likelihood estimation and a minimum entropy estimation for the expected value and variance of normal fuzzy variable are discussed within the framework of credibility theory. As an application, a credibilistic portfolio selection model is proposed, which is an improvement over the traditional models as it only needs the predicted values on the security returns instead of...
متن کاملModeling of the Maximum Entropy Problem as an Optimal Control Problem and its Application to Pdf Estimation of Electricity Price
In this paper, the continuous optimal control theory is used to model and solve the maximum entropy problem for a continuous random variable. The maximum entropy principle provides a method to obtain least-biased probability density function (Pdf) estimation. In this paper, to find a closed form solution for the maximum entropy problem with any number of moment constraints, the entropy is consi...
متن کاملDiscriminative maximum entropy language model for speech recognition
This paper presents a new discriminative language model based on the whole-sentence maximum entropy (ME) framework. In the proposed discriminative ME (DME) model, we exploit an integrated linguistic and acoustic model, which properly incorporates the features from n-gram model and acoustic log likelihoods of target and competing models. Through the constrained optimization of integrated model, ...
متن کاملContext-dependent acoustic modeling based on hidden maximum entropy model for statistical parametric speech synthesis
semi Markov models (HSMMs) are typically used in statistical parametric speech synthesis to represent probability densities of acoustic features given contextual factors. This paper addresses three major limitations of this decision tree-based structure: i) the decision tree structure lacks adequate context generalization; ii) it is unable to express complex context dependencies; iii) parameter...
متن کاملE-Bayesian Approach in A Shrinkage Estimation of Parameter of Inverse Rayleigh Distribution under General Entropy Loss Function
Whenever approximate and initial information about the unknown parameter of a distribution is available, the shrinkage estimation method can be used to estimate it. In this paper, first the $ E $-Bayesian estimation of the parameter of inverse Rayleigh distribution under the general entropy loss function is obtained. Then, the shrinkage estimate of the inverse Rayleigh distribution parameter i...
متن کامل