Speech synthesis systems still fail in producing acceptable prosodies. We are developing a research strategy designed to de-focus attention on the objective acoustic accuracy of synthetic speech in favour of enhancing the speech to optimize a listener’s ability to repair ‘damaged’ signals. To do this we need to know more about how listeners repair errors and how we might trigger the repair proc...

Displayless interface technology provides speech-based access to computer applications for which visual access is not possible. These applications are increasingly prevalent, especially in situations requiring mobility, such as navigational applications. To ensure the successful deployment of this technology however, many human factors issues must be addressed. In particular, its nonvisual natu...

The aim of the first joint Speech and Natural Language Workshop was to bring together these two research communities, to interchange technical information, to reflect on past successes and to define future directions for Spoken Language research. The overall organization of the technical program was designed 1) to establish some common reference points by assessing the current state of the a r ...

In this paper we discuss some issues in processing speech signals, especially for isolated utterances of characters of a language. For processing this speech signal we have no clues of higher level linguistic information such a s prosodics, lexical, syntax, and semantics. Any representation of signals in terms of fixed parameters for each short (10-20 msec) segment is not likely to provide the ...

Most current state-of-the-art automatic speaker recognition systems extract speaker-dependent features by looking at shortterm spectral information. This approach ignores long-term information that can convey supra-segmental information, such as prosodics and speaking style. We propose two approaches that use the fundamental frequency and energy trajectories to capture long-term information. Th...

Text-independent speaker recognition systems such as those based on Gaussian mixture models (GMMs) do not include time sequence information (TSI) within the model itself. The level of importance of TSI in speaker recognition is an interesting question and one addressed in this paper. Recent works has shown that the utilisation of higher-level information such as idiolect, pronunciation, and pro...

The usefulness of teaching pronunciation in language instruction remains controversial. Though past research suggests that teachers can make little or no difference in improving their students’ pronunciation,  current  findings  suggest  that  second  language  pronunciation  can  improve  to  be near  native-like  with  the  implementation  of  certain  criteria  such  as  the  utilization  of...

This paper describes a preliminary work on prosody modeling aspect of a spoken language understanding system for Thai. Specifically, the model is designed to integrate prosodics into a language model based on constraint dependency grammar. There are two steps involved, namely the prosodic annotation process and the prosodic disambiguation process. The annotation process uses prosodic informatio...

The speech production system is capable of conveying an abundance of information with regards to sentence text, speaker identity, prosodics, as well as emotion and speaker stress. In an effort to better understand the mechanism of human voice communication, researchers have attempted to determine reliable acoustic indicators of stress using such speech production features as fundamental frequen...