Toward the border between neural and Markovian paradigms

Wilinski, P.; Solaiman, Basel; Hillion, A.; Czarnecki, Wojciech Marian

doi:10.1109/3477.662756

Cited by 11 publications

(8 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the early 1990s this fact led to the idea of combining HMMs and ANNs within a single, novel model, broadly known as hybrid HMM/ANN [38,70,15,87,93,44,118,8]. A recently published survey paper [125] investigates the`bordera between ANNs and Markovian models (HMMs for ASR and Markov random xelds for image processing), summarizing approaches where ANNs emulate Markov models and vice versa, and reporting hybrid architectures. The hybrid paradigm relies on maintaining an underlying HMM structure, capable of modeling longterm dependencies, with the integration of ANNs, which provide non-parametric universal approximation, probability estimation, discriminative training algorithms, fewer parameters to estimate than those usually required in standard HMMs, e$cient computation of outputs at recognition time, and e$cient hardware implementability.…”

Section: Introductionmentioning

confidence: 99%

A survey of hybrid ANN/HMM models for automatic speech recognition

2001

View full text Add to dashboard Cite

In spite of the advances accomplished throughout the last decades, automatic speech recognition (ASR) is still a challenging and di$cult task. In particular, recognition systems based on hidden Markov models (HMMs) are e!ective under many circumstances, but do su!er from some major limitations that limit applicability of ASR technology in real-world environments. Attempts were made to overcome these limitations with the adoption of arti"cial neural networks (ANN) as an alternative paradigm for ASR, but ANN were unsuccessful in dealing with long time-sequences of speech signals. Between the end of the 1980s and the beginning of the 1990s, some researchers began exploring a new research area, by combining HMMs and ANNs within a single, hybrid architecture. The goal in hybrid systems for ASR is to take advantage from the properties of both HMMs and ANNs, improving #exibility and recognition performance. A variety of di!erent architectures and novel training algorithms have been proposed in literature. This paper reviews a number of signi"cant hybrid models for ASR, putting together approaches and techniques from a highly specialistic and non-homogeneous literature. E!orts concentrate on describing and referencing architectures and algorithms, their advantages and limitations, as well as on categorizing them into broad classes. Early attempts to emulate HMMs by ANNs are "rst described. Then we focus on ANNs to estimate posterior probabilities of the states of an HMM and on`globala optimization, where a single, overall training criterion is de"ned over the HMM and the ANNs. Connectionist vector quantization for discrete HMMs, and other more recent approaches are also reviewed. It is pointed out that, in addition to their theoretical interest, hybrid systems have been allowing for tangible improvements in recognition performance over the standard HMMs in di$cult and signi"cant benchmark tasks.

show abstract

Section: Introductionmentioning

confidence: 99%

A survey of hybrid ANN/HMM models for automatic speech recognition

2001

View full text Add to dashboard Cite

show abstract

“…Exactly like in (6), we can use gradient ascent to maximize by taking (23) for a generic weight . From (22)…”

Section: B Imposing a Probabilistic Constraint On The Ann Weightsmentioning

confidence: 99%

“…A number of significant, different hybrid paradigms for ASR were proposed in the literature, e.g., [14]- [22]. Surveys on the topic can be found in [5], [23]. The approach presented in this paper is related to some extent to Bourlard and Morgan's hybrid [4], [24], [25], as well as to Bengio's optimization scheme [26].…”

Section: Introductionmentioning

confidence: 99%

Robust combination of neural networks and hidden markov models for speech recognition

Trentin

Gori

2003

IEEE Trans. Neural Netw.

View full text Add to dashboard Cite

Acoustic modeling in state-of-the-art speech recognition systems usually relies on hidden Markov models (HMMs) with Gaussian emission densities. HMMs suffer from intrinsic limitations, mainly due to their arbitrary parametric assumption. Artificial neural networks (ANNs) appear to be a promising alternative in this respect, but they historically failed as a general solution to the acoustic modeling problem. This paper introduces algorithms based on a gradient-ascent technique for global training of a hybrid ANN/HMM system, in which the ANN is trained for estimating the emission probabilities of the states of the HMM. The approach is related to the major hybrid systems proposed by Bourlard and Morgan and by Bengio, with the aim of combining their benefits within a unified framework and to overcome their limitations. Several viable solutions to the "divergence problem"-that may arise when training is accomplished over the maximum-likelihood (ML) criterion-are proposed. Experimental results in speaker-independent, continuous speech recognition over Italian digit-strings validate the novel hybrid framework, allowing for improved recognition performance over HMMs with mixtures of Gaussian components, as well as over Bourlard and Morgan's paradigm. In particular, it is shown that the maximum a posteriori (MAP) version of the algorithm yields a 46.34% relative word error rate reduction with respect to standard HMMs.

show abstract

“…Because individual methods are not perfect for attaining the goal, two or more methodologies are combined in a way that their weaknesses are compensated by each other, and advantages of both methodologies are used to achieve the solution. Earlier approaches to such models are summarized in Willinski et al (1998). In this research, both the methods are distinct, and only intermediary results are exchanged.…”

Section: Introductionmentioning

confidence: 99%

“…In this research, both the methods are distinct, and only intermediary results are exchanged. Earlier approaches to such models are summarized in Willinski et al (1998). The authors have used ANNs and FISs as the soft computing methods for the prediction of time series.…”

Section: Introductionmentioning

confidence: 99%

Pattern‐Similarity‐Based Model for Time Series Prediction

Bhardwaj

Srivastava

Gupta

2013

Computational Intelligence

View full text Add to dashboard Cite

This research proposes a pattern/shape-similarity-based clustering approach for time series prediction. This article uses single hidden Markov model (HMM) for clustering and combines it with soft computing techniques (fuzzy inference system/artificial neural network) for the prediction of time series. Instead of using distance function as an index of similarity, here shape/pattern of the sequence is used as the similarity index for clustering, which overcomes few of the shortcomings associated with distance-based clustering approaches. Underlying hidden properties of time series are captured with the help of HMM. The prediction method used here exploits the pattern identification prowess of the HMM for cluster selection and the generalization and nonlinear modeling capabilities of soft computing methods to predict the output of the system. To see the validity of the proposed method in the real-life scenario, it is tested on four different time series. The first is a benchmark Mackey-Glass time series, which is tested for delay parameters D 17 and D 30. The remaining time series are monthly sunspot data time series, Laser data time series and the last is Lorenz attractor time series. Simulation results show that the proposed method provide a better prediction performance in comparison with the existing methods.

show abstract

Toward the border between neural and Markovian paradigms

Cited by 11 publications

References 64 publications

A survey of hybrid ANN/HMM models for automatic speech recognition

A survey of hybrid ANN/HMM models for automatic speech recognition

Robust combination of neural networks and hidden markov models for speech recognition

Pattern‐Similarity‐Based Model for Time Series Prediction

Contact Info

Product

Resources

About