Automatic transcription of conversational telephone speech

This paper discusses the Cambridge University HTK (CU-HTK) system for the automatic transcription of conversational telephone speech. A detailed discussion of the most important techniques in front-end processing, acoustic modeling and model training, language and pronunciation modeling are presented. These include the use of conversation side based cepstral normalization, vocal tract length normalization, heteroscedastic linear discriminant analysis for feature projection, minimum phone error training and speaker adaptive training, lattice-based model adaptation, confusion network based decoding and confidence score estimation, pronunciation selection, language model interpolation, and class based language models. The transcription system developed for participation in the 2002 NIST Rich Transcription evaluations of English conversational telephone speech data is presented in detail. In this evaluation the CU-HTK system gave an overall word error rate of 23.9%, which was the best performance by a statistically significant margin. Further details on the derivation of faster systems with moderate performance degradation are discussed in the context of the 2002 CU-HTK 10 × RT conversational speech transcription system. © 2005 IEEE.

Unsupervised training for mandarin broadcast news and conversation transcription

A significant cost in obtaining acoustic training data is the generation of accurate transcriptions. For some sources close-caption data is available. This allows the use of lightly-supervised training techniques. However, for some sources and languages close-caption is not available. In these cases unsupervised training techniques must be used. This paper examines the use of unsupervised techniques for discriminative training. In unsupervised training automatic transcriptions from a recognition system are used for training. As these transcriptions may be errorful data selection may be useful. Two forms of selection are described, one to remove non-target language shows, the other to remove segments with low confidence. Experiments were carried out on a Mandarin transcriptions task. Two types of test data were considered, Broadcast News (BN) and Broadcast Conversations (BC). Results show that the gains from unsupervised discriminative training are highly dependent on the accuracy of the automatic transcriptions. © 2007 IEEE.

Simulating dynamic covariance structures for testing the adaptive behavior of variable selection algorithms (Invited Paper)

Variable selection for regression is a classical statistical problem, motivated by concerns that too large a number of covariates may bring about overfitting and unnecessarily high measurement costs. Novel difficulties arise in streaming contexts, where the correlation structure of the process may be drifting, in which case it must be constantly tracked so that selections may be revised accordingly. A particularly interesting phenomenon is that non-selected covariates become missing variables, inducing bias on subsequent decisions. This raises an intricate exploration-exploitation tradeoff, whose dependence on the covariance tracking algorithm and the choice of variable selection scheme is too complex to be dealt with analytically. We hence capitalise on the strength of simulations to explore this problem, taking the opportunity to tackle the difficult task of simulating dynamic correlation structures. © 2008 IEEE.

Streaming covariance selection with applications to adaptive querying in sensor networks

Sensor networks can be naturally represented as graphical models, where the edge set encodes the presence of sparsity in the correlation structure between sensors. Such graphical representations can be valuable for information mining purposes as well as for optimizing bandwidth and battery usage with minimal loss of estimation accuracy. We use a computationally efficient technique for estimating sparse graphical models which fits a sparse linear regression locally at each node of the graph via the Lasso estimator. Using a recently suggested online, temporally adaptive implementation of the Lasso, we propose an algorithm for streaming graphical model selection over sensor networks. With battery consumption minimization applications in mind, we use this algorithm as the basis of an adaptive querying scheme. We discuss implementation issues in the context of environmental monitoring using sensor networks, where the objective is short-term forecasting of local wind direction. The algorithm is tested against real UK weather data and conclusions are drawn about certain tradeoffs inherent in decentralized sensor networks data analysis. © 2010 The Author. Published by Oxford University Press on behalf of The British Computer Society. All rights reserved.

A new method for the acquisition of ultrasonic strain image volumes.

This article presents a new method for acquiring three-dimensional (3-D) volumes of ultrasonic axial strain data. The method uses a mechanically-swept probe to sweep out a single volume while applying a continuously varying axial compression. Acquisition of a volume takes 15-20 s. A strain volume is then calculated by comparing frame pairs throughout the sequence. The method uses strain quality estimates to automatically pick out high quality frame pairs, and so does not require careful control of the axial compression. In a series of in vitro and in vivo experiments, we quantify the image quality of the new method and also assess its ease of use. Results are compared with those for the current best alternative, which calculates strain between two complete volumes. The volume pair approach can produce high quality data, but skillful scanning is required to acquire two volumes with appropriate relative strain. In the new method, the automatic quality-weighted selection of image pairs overcomes this difficulty and the method produces superior quality images with a relatively relaxed scanning technique.

Bayesian model selection for time series using Markov chain Monte Carlo

We present a stochastic simulation technique for subset selection in time series models, based on the use of indicator variables with the Gibbs sampler within a hierarchical Bayesian framework. As an example, the method is applied to the selection of subset linear AR models, in which only significant lags are included. Joint sampling of the indicators and parameters is found to speed convergence. We discuss the possibility of model mixing where the model is not well determined by the data, and the extension of the approach to include non-linear model terms.