Learn More
This paper introduces a new metric for the quantitative assessment of the similarity of speakers' accents. The ACCDIST metric is based on the correlation of inter-segment distance tables across speakers or groups. Basing the metric on segment similarity within a speaker ensures that it is sensitive to the speaker's pronunciation system rather than to his or(More)
We have developed a novel therapy based on a computer program, which enables the patient to create an avatar of the entity, human or non-human, which they believe is persecuting them. The therapist encourages the patient to enter into a dialogue with their avatar, and is able to use the program to change the avatar so that it comes under the patient's(More)
This paper describes the results of a study of the phonetic and phonological factors affecting the rhythm and timing of spoken Korean. Stepwise construction of a CART model was used to uncover the contribution and relative importance of phrasal, syllabic, and segmental contexts. The model was trained from a corpus of 671 read sentences, yielding 42,000(More)
We propose a data driven, non-intrusive method for speech intelligibility estimation. We begin with a large set of speech signal specific features and use a dimensionality reduction approach based on correlation and principal component analysis to find the most relevant features for intelligibility prediction. These are then used to train a Gaussian mixture(More)
Speech synthesis research has been transformed in recent years through the exploitation of speech corpora – both for statistical modelling and as a source of signals for concatenative synthesis. This revolution in methodology and the new techniques it brings calls into question the received wisdom that better computer voice output will come from a better(More)
Intonation modelling in ProSynth involves mapping the defining characteristics of an F0 contour on to the constituents of a hierarchical prosodic structure, which constitutes our core linguistic representation. The paper describes the use of a labelled speech database exemplifying selected structures to create a template for a particular pitch pattern in a(More)
SPAR (Speech-Pattern Algorithms and Representations) is the name given to The project is concerned with advanced speech analysis algorithms, and from the outset saw the need for a system for speech data management. The SPAR Speech Filing System (SFS) was developed to support the design and comparison of analysis algorithms, and to manage many different(More)