An Articulatory Feature-Based Tandem Approach and Factored Observation Modeling

Çetin, Özgür; Kantor, Arthur; King, Simon; Bartels, Chris; Magimai-Doss, Mathew; Frankel, Joe; Livescu, Karen

doi:10.1109/icassp.2007.366995

Cited by 40 publications

(44 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These are not exclusive; many systems often concatenate both acoustic features and posteriors derived from them as inputs to a system (e.g., [9]). • Statistical model: HMMs correspond to a family of generative models; that is, the statistical models are formulated so that they model data generation, requiring a likelihood of the data as well as a model prior to compute model posteriors.…”

Section: Prior Workmentioning

confidence: 99%

“…In addition to the Hybrid system, [4] also describes a system that uses outputs of a set of phonological attribute classifiers in an HMM system; this style of system has been further explored in [8] and in [9]. 3 As the Tandem HMM system is built using mixtures of Gaussians to describe state emission probabilities, either the correlated phonological feature inputs must first be decorrelated before being fed into the system or the system must make use of full or semi-tied covariance matrices and suffer an explosion in parameters and required training data.…”

Section: Prior Workmentioning

confidence: 99%

See 1 more Smart Citation

Conditional Random Fields for Integrating Local Discriminative Classifiers

Morris

Fosler‐Lussier

2008

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-Conditional random fields (CRFs) are a statistical framework that has recently gained in popularity in both the automatic speech recognition (ASR) and natural language processing communities because of the different nature of assumptions that are made in predicting sequences of labels compared to the more traditional hidden Markov model (HMM). In the ASR community, CRFs have been employed in a method similar to that of HMMs, using the sufficient statistics of input data to compute the probability of label sequences given acoustic input. In this paper, we explore the application of CRFs to combine local posterior estimates provided by multilayer perceptrons (MLPs) corresponding to the frame-level prediction of phone classes and phonological attribute classes. We compare phonetic recognition using CRFs to an HMM system trained on the same input features and show that the monophone label CRF is able to achieve superior performance to a monophone-based HMM and performance comparable to a 16 Gaussian mixture triphone-based HMM; in both of these cases, the CRF obtains these results with far fewer free parameters. The CRF is also able to better combine these posterior estimators, achieving a substantial increase in performance over an HMM-based triphone system by mixing the two highly correlated sets of phone class and phonetic attribute class posteriors.

show abstract

Section: Prior Workmentioning

confidence: 99%

Section: Prior Workmentioning

confidence: 99%

Conditional Random Fields for Integrating Local Discriminative Classifiers

Morris

Fosler‐Lussier

2008

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…The 500 word baseline system has a small improvement over the results presented in [13]. Note that systems that train with additional data outside the designated SVitchboard training sets have reported lower word error rates [13,20].…”

Section: Resultsmentioning

confidence: 82%

Use of syllable nuclei locations to improve ASR

Bartels

Bilmes

2007

2007 IEEE Workshop on Automatic Speech Recognition &Amp; Understanding (ASRU)

View full text Add to dashboard Cite

This work presents the use of dynamic Bayesian networks (DBNs) to jointly estimate word position and word identity in an automatic speech recognition system. In particular, we have augmented a standard Hidden Markov Model (HMM) with counts and locations of syllable nuclei. Three experiments are presented here. The first uses oracle syllable counts, the second uses oracle syllable nuclei locations, and the third uses estimated (non-oracle) syllable nuclei locations. All results are presented on the 10 and 500 word tasks of the SVitchboard corpus. The oracle experiments give relative improvements ranging from 7.0% to 37.2%. When using estimated syllable nuclei a relative improvement of 3.1% is obtained on the 10 word task.

show abstract

“…Most tandem features to date have relied on MLPs trained for phone classification. Cetin et al [105] illustrated on a relatively small data set that MLPs trained for articulatory feature classification can be equally effective. They provided a similar comparison using MLPs trained on a much larger data set -2,000 h of English conversational telephone speech.…”

Section: Speech and Audio Processingmentioning

confidence: 99%

Computational Intelligence in Multimedia Processing: Foundation and Trends

Hassanien

Abraham²,

Kacprzyk³

et al. 2008

Studies in Computational Intelligence

View full text Add to dashboard Cite

Summary. This chapter presents a broad overview of Computational Intelligence (CI) techniques including Neural Network (NN), Particle Swarm Optimization (PSO), Evolutionary Algorithm (GA), Fuzzy Set (FS), and Rough Sets (RS). In addition, a very brief introduction to near sets and near images which offer a generalization of traditional rough set theory and a new approach to classifying perceptual objects by means of features in solving multimedia problems is presented. A review of the current literature on CI based approaches to various problems in multimedia computing such as speech, audio and image processing, video watermarking, content-based multimedia indexing and retrieval are presented. We discuss some representative methods to provide inspiring examples to illustrate how CI could be applied to resolve multimedia computing problems and how multimedia could be analyzed, processed, and characterized by computational intelligence. Challenges to be addressed and future directions of research are also presented.

show abstract

An Articulatory Feature-Based Tandem Approach and Factored Observation Modeling

Cited by 40 publications

References 9 publications

Conditional Random Fields for Integrating Local Discriminative Classifiers

Conditional Random Fields for Integrating Local Discriminative Classifiers

Use of syllable nuclei locations to improve ASR

Computational Intelligence in Multimedia Processing: Foundation and Trends

Contact Info

Product

Resources

About