B. H. Juang and L. R. Rabiner. Hidden Markov models for speech recognition. Technometrics, 33:251--272, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and Basic Definitions 1.1 Overview Hidden Markov Models (HMMs) are one of the more popular and successful techniques for pattern recognition in use today.For example, experiments in speech recognition haveshown that HMMs can be useful tools in modelling the variabilityofhuman speech. [juang91], lee88] rabiner86] bahl88] Hidden Markov Models have also been used in computational linguistics [kupiec90] in document recognition [kopec91] and in such situations where intrinsic statistical variabilityindatamust be accounted for in order to perform pattern recognition. HMMs are constructed ....

B.H. Juang and L.R.Rabiner, Hidden Markov Models for Speech Recognition, Technometrics, Vol.33, No.3, August 1991.

....The well known Viterbi algorithm nds the sequence of states that is most likely to have generated a given observation sequence. The Baum Welch algorithm, an instantiation of EM, can be used to estimate the most likely HMM parameters given a collection of observation sequences. Speech recognition [11] and computational biochemistry [1] are well known applications of HMMs. Non hidden) Markov model algorithms that are used for part of speech tagging [3] and for information extraction [14] require each observation (i.e. token) of the observation sequences (documents) used for training to be ....

B. Juang and L. Rabiner. Hidden markov models for speech recognition. Technometrics, 33:251{ 272, 1991.

....point of view, we also believe that it may be beneficial to give up local normalization of parameters even for standard HMMs. In fact, non normalizing parameters are already used frequently in speech recognition by introducing so called transition biases and stream exponents , see e.g. [6, 8, 13]. These heuristic approaches are used in order to reduce the mismatch between transition and emission probabilities in standard HMMs. In [10, 13] these issues are discussed in greater detail. 2.1. Training and decoding To train the model we assume that the complete labeling is available ....

.... 94, pp. 239 42, 1994. 6] F. Johansen and M. Johnsen, Non linear input transformations for discriminative HMMs, in Proceedings of ICASSP 94, pp. 225 28, 1994. 7] B. Juang and L. Rabiner, Hidden Markov models for speech recognition, Technometrics, vol. 33, no. 3, pp. 251 72, 1991. [8] S. Kapadia, V. Valtchev, and S. Young, Mmi training for continuous phoneme recognition on the timit database, in Proceedings of ICASSP 93, pp. 491 4, 1993. 9] A. Krogh, Hidden Markov models for labeled sequences, in Proceedings of the 12th ICPR 94, pp. 140 4, 1994. 10] A. Krogh and ....

[Article contains additional citation context not shown here]

JUANG, B. H., AND RABINER, L. R. Hidden Markov models for speech recognition. Technometrics 33, 3 (August 1991), 251--272.

....under given model parameters. The basic theory of the HMMs was presented already at 60 s and 70 s by Baum et al. 4] 8] The rst applications were made in speech recognition [3] In last ten years a huge amount of applications of HMMs have been published in the eld of speech recognition [2] [18], 24] An especially useful and accurate tutorial has been written by Rabiner [27] In addition to the speech recognition, HMMs have been widely used for other types of pattern recognition, e.g. ECG signal recognition [21] and text recognition [23] 19] In the area of molecular biology, HMMs ....

Juang, B. H., and Rabiner, L. R. (1991), "Hidden Markov Models for Speech Recognition", Technometrics, 33, 251-272.

....and hence all the conclusions about the process must be made using only the observation sequence. The basic theory of HMMs was introduced already in the late 60 s by Baum and Petrie (1966) and thereafter the model has been extensively used in many areas such as speech processing (Baker, 1975; Juang and Rabiner, 1991), recognition of handwritten word (Kundu et al. 1989) and modeling and analysis of DNA and protein sequences (Churchill, 1989; Eddy, 1998) In the applications of HMMs, likelihood functions and estimates of the model parameters have been routinely computed. However, the more advanced statistical ....

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition, Technometrics, 33, 251-272.

....P (Y t jS t ) can be fully specified as a K Theta L observation matrix. For a continuous observation vector, P (Y t jS t ) can be modeled in many different forms, such as a Gaussian, mixture of Gaussians, or a neural network. HMMs have been applied extensively to problems in speech recognition (Juang and Rabiner, 1991), computational biology (Baldi et al. 1994) and fault detection (Smyth, 1994) Given an HMM with known parameters and a sequence of observations, two algorithms are commonly used to solve two different forms of the inference problem (Rabiner and Juang, 1986) The first computes the posterior ....

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition. Technometrics, 33:251--272.

....and not merely as a computational device. The posterior distribution retains the Markov property, conditional on the data, and can be simulated via the backward recursion in Baum, Petrie, Soules and Weiss (1970) Applications of hidden Markov modes occur in speech recognition (e.g. Rabiner, 1989; Juang and Rabiner, 1991), in neurophysiology (e.g. Fredkin and Rice, 1992) in computational biology (e.g. Haussler, Krogh, Mian and Sjolander, 1993; Eddie, Mitchison and Durbin, 1995; Liu, Neuwald and Lawrence, 1995) in climatology (e.g. Hughes, Guttorp and Charles, 1999) in epidemiologic surveillance (Le Strat and ....

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition. Technometrics, 33, 251-272.

....and Basic Definitions 1.1 Overview Hidden Markov Models (HMMs) are one of the more popular and successful techniques for pattern recognition in use today. For example, experiments in speech recognition have shown that HMMs can be useful tools in modelling the variability of human speech. [juang91], lee88] rabiner86] bahl88] Hidden Markov Models have also been used in computational linguistics [kupiec90] in document recognition [kopec91] and in such situations where intrinsic statistical variability in data must be accounted for in order to perform pattern recognition. HMMs are ....

B.H. Juang and L.R.Rabiner, Hidden Markov Models for Speech Recognition, Technometrics, Vol.33, No.3, August 1991.

.... accuracy we choose parameters so as to maximize, P (yjx; M) p(x; yjM) p(xjM) 4) as we have previously proposed in [9] where it was called CHMM for class HMM ) This has also been called Conditional Maximum Likelihood (CML) and is equivalent to Maximum Mutual Information estimation (MMI) [1, 7] if the language model is fixed. p(x; yjM) is calculated as a sum over all paths consistent with the labeling, i.e. if observation l is labeled by f only paths in which the l th state has label f are allowed. If the set of these consistent paths is called A(y) we have, p(x; yjM) q(x; yjM) ....

....forward algorithm, see [9, 10] 1 In speech recognition this is not always the case. Often only the sequence of utterance symbols, e.g. the phoneme transcription, is known (incomplete labeling) However, the framework presented here can easily be adapted to the case of incomplete labeling, see [7, 10]. To optimize (8) we use gradient descent. Calculating the derivative of log P (yjx; M) w. r. t. a weight in the match or transition networks, yields backpropagation training of the neural networks based on an error signal calculated by the forward backward algorithm, see [10] for details. This ....

[Article contains additional citation context not shown here]

B. Juang and L. Rabiner, "Hidden Markov models for speech recognition," Technometrics, vol. 33, no. 3, pp. 251--72, 1991.

....on the task of recognizing broad phoneme classes in the TIMIT database shows clear performance gains compared to standard HMMs tested on the same task. 1 Introduction Hidden Markov models is one of the most successful modeling approaches for acoustic events in speech recognition (Rabiner 1989; Juang Rabiner 1991), and more recently they have proven useful for several problems in biological sequence analysis like protein modeling and gene finding, see e.g. Durbin et al. 1998; Eddy 1996; Krogh et al. 1994) Although the HMM is good at capturing the temporal nature of processes such as speech it has a very ....

....more powerful model with better classification abilities. The starting point for this work is the so called class HMM (CHMM) which is basically a standard HMM with a distribution over classes assigned to each state (Krogh 1994) The CHMM incorporates conditional maximum likelihood (CML) estimation (Juang Rabiner 1991; N adas 1983; N adas, Nahamoo, Picheny 1988) In contrast to the widely used Maximum Likelihood (ML) estimation, CML estimation is a discriminative training algorithm that aims at maximizing the ability of the model to discriminate between different classes. The CHMM can be normalized globally, ....

[Article contains additional citation context not shown here]

Juang, B. H., and Rabiner, L. R. 1991. Hidden Markov models for speech recognition.

....P (Y t jS t ) can be fully specified as a K Theta L observation matrix. For a continuous observation vector, P (Y t jS t ) can be modeled in many different forms, such as a Gaussian, mixture of Gaussians, or a neural network. HMMs have been applied extensively to problems in speech recognition (Juang and Rabiner, 1991), computational biology (Baldi et al. 1994) and fault detection (Smyth, 1994) Given an HMM with known parameters and a sequence of observations, two algorithms are commonly used to solve two different forms of the inference problem (Rabiner and Juang, 1986) The first computes the posterior ....

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition. Technometrics, 33:251--272.

.... nous int eresser a ce probl eme, signalons que le mod ele de chaines de Markov cach ees (CMC) dont la premi ere etude significative est due a Baum, Petrie, Soulie et Weiss [BAU 70] est devenu un mod ele des plus classiques en reconnaissance de la parole (cf. par exemple, RAB 85] GUC 90] JUR 91] ou [AND 92] et, pour un etat de l art r ecent [GUE 92] A l heure actuelle, des applications dans de nombreux domaines se multiplient : donn ees neurophysiologiques ( ALB 91] LLP 92] analyse de la structure de l ADN ( CHU 89] Quant au pr esent article, il contient une application a ....

....de Viterbi Cet al..gorithme est l algorithme du MAP adapt e au mod ele des CMC. On d ecrit ainsi l it eration m de cet al..gorithme : Etape C : l etape de restauration des donn ees manquantes est une etape de classification. On construit z m par le principe du MAP. Cela donne (cf. par exemple, JUR 91] z m i = arg max =1; k t m z m i Gamma1 OE(x i j m )t m z m Gamma1 i 1 : Etape M : les termes de la matrice D m 1 sont calcul es par la formule D m 1 j = 1 n n X i=2 I fz m i Gamma1 =j; z m i = g De l a, on d eduit ais ement l actualisation de la matrice de ....

JUANG, B.H. et RABINER, L.R. : Hidden Markov Models for Speech Recognition. Technometrics 33, 251-272, 1991.

....P (Y t jS t ) can be fully speci ed as a K L observation matrix. For a continuous observation vector, P (Y t jS t ) can be modeled in many di erent forms, such as a Gaussian, mixture of Gaussians, or a neural network. HMMs have been applied extensively to problems in speech recognition (Juang and Rabiner, 1991), computational biology (Baldi et al. 1994) and fault detection (Smyth, 1994) Given an HMM with known parameters and a sequence of observations, two algorithms are commonly used to solve two di erent forms of the inference problem (Rabiner and Juang, 1986) The rst computes the posterior ....

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition. Technometrics, 33:251-272.

.... probability of the correct labeling y associated with observation sequence x, P (yjx; M) P (x; yjM) P (xjM) 1) as we have previously proposed in [10] Maximizing (1) is known as Conditional Maximum Likelihood estimation (CML) and is equivalent to Maximum Mutual Information estimation (MMI) [1, 8] if the language model is fixed during training. For an observation sequence of length L, the labeling y can be either complete, i.e. there is one label for each observation (y = y1 ; yL ) or incomplete, i.e. the label sequence y = y1 ; yS is shorter than the observation ....

JUANG, B. H., AND RABINER, L. R. Hidden Markov models for speech recognition. Technometrics 33, 3 (August 1991), 251--272.

....by exact computations and only the remaining, intractable parts of the network are handled within mean field theory. For simplicity we focus on networks with binary units; the extension to discrete valued (Potts) units is straightforward. We apply these ideas to hidden Markov modeling (Rabiner Juang, 1991). The first order probabilistic structure of hidden Markov models (HMMs) leads to networks with chained architectures for which efficient, exact algorithms are available. More elaborate networks are obtained by introducing couplings between multiple HMMs (Williams Hinton, 1990) and or long range ....

B. H. Juang and L. R. Rabiner. (1991) Hidden Markov models for speech recognition, Technometrics 33: 251--272.

No context found.

B. H. Juang and L. R. Rabiner. Hidden Markov models for speech recognition. Technometrics, 33:251--272, 1991.

No context found.

Juang, B. H. and Rabiner, L. R. (1991). Hidden Markov models for speech recognition.

No context found.

B. Juang and L. Rabiner. Hidden markov models for speech recognition. Technometrics, 33:251{ 272, 1991.

No context found.

B. H. Juang and L. R. Rabiner. Hidden Markov models for speech recognition. Technometrics, 33:251-- 272, 1991.

No context found.

B. H. Juang and L. R. Rabiner, \Hidden Markov models in speech recognition," Technometrics, vol. 33, pp. 251-272, 1991.

No context found.

B.H. Juang and L.R. Rabiner Hidden Markov models for speech recognition. Technometrics, vol 33 , 251272, 1991.

No context found.

Juang, B.H. and Rabiner, L.R. (1990), "Hidden Markov models for speech recognition," Technometrics, 33, 251-272.

No context found.

Juang, B.H. and Rabiner, L.R. (1991) Hidden Markov models for speech recognition.

No context found.

B. H. Juang and L. R. Rabiner, "Hidden markov models for speech recognition," Technometrics, vol. 33, pp. 251--272, August 1991.

No context found.

Juang, B.H. and L.R. Rabiner (1991) Hidden Markov models for speech recognition. Technometrics, 33, 251-272.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC