Showing papers in "Computer Speech & Language in 1993"

TL;DR: This work proposes an alternative representation of hidden Markov models in which a state of an HMM is defined as a template, i.e. a "typical" sequence of observations, derived from an ensemble of segments corresponding to that state.

TL;DR: Perceptual methods for diagnosing problems in text-to-speech systems are described and a battery of experimental paradigms that address different facets of speech quality and intelligibility are discussed.

TL;DR: Two scoring algorithms to rank candidate parses are proposed, both based on an analysis/synthesis approach that compares the recognized prosodic phrase structure (analysis) with the predicted structure (Synthesis) for each candidate parse.

TL;DR: The dimension of the frame feature vectors, and hence the number of model parameters, were greatly reduced without a significant loss of recognition performance.

TL;DR: The goal of this paper is to introduce and describe a formalism for segment based phonology and phonological processing.

TL;DR: The mu+ system has been developed to provide a common environment for experimentation in numerous facets of corpus based speech and language research including: articulatory and acoustic phonetics, prosodic analysis, speech technology research, and linguistic corpus development.

TL;DR: It is suggested that the decomposition of the utterance into intonation phrases by means of pitch makes an essential contribution to the naturalness of synthetic speech.

TL;DR: The semicontinuous hidden Markov model was extended to incorporate multiple code-books and it was found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability.

TL;DR: Speech recognition results show that the new system outperforms the traditional HMM approaches in small tasks and Examination of the source of error, using Viterbi analysis, suggests that this new scheme is able to achieve better modelling of the acoustic transitions and coarticulation in speech.

TL;DR: Some features of the fundamental frequency (F0) contours of speech in Hindi are described and an approach to represent and activate this intonation knowledge for an unrestricted text-to-speech system for Hindi is proposed.

TL;DR: In this paper, the authors used the information theoretic measure of mutual information to investigate the distribution of phonetic information across the on/off aligned auditory spectrogram for a corpus of vowel-plosive-vowel utterances.

TL;DR: The performance of continuous HMMs using one type of transitional features in speaker-dependent recognition of the highly confusing Mandarin syllables is first evaluated and discussed in detail under the constraint of very limited training data.

TL;DR: A text-to-speech system for Dutch, called Spraakmaker, is described, based on a flexible underlying framework which has been devised for buildingText Maker, which is a multi-level, synchronized data structure based on the work of Hertz, Kadin and Karplus (1985), which is to contain all linguistic information relevant to the text- to-speech conversion process.

TL;DR: This paper shows how the improved acoustic modeling techniques (using a continuous density hidden Markov model framework), developed for large-vocabulary speech recognition applications, can be applied to the problem of connected digit recognition with no changes made to the basic modeling techniques and with no vocabulary-specific information used.

TL;DR: It is concluded that the phoneme detection task is a useful tool for investigating phonetic processing of synthetic speech input, but subjects must be encouraged to adopt a response criterion which emphasizes rapid responding.

TL;DR: The evidence indicates that the process of formalizing spectrogram reading can be modeled with rules, and the knowledge acquisition and knowledge representation, in terms of descriptions and rules, are described.

TL;DR: An extensive and detailed description is given of the possibilities of SMF, concerning both the specification of patterns in the data structure and the alteration of its contents, and the completeness and the expressive power of the formalism are discussed.