Loading...
Please wait, while we are loading the content...
Similar Documents
Building Synthetic Voice : A Journey from the Known Voices to a Kiwi Voice
| Content Provider | Semantic Scholar |
|---|---|
| Copyright Year | 2004 |
| Abstract | This paper relates to a work in progress and is an offshoot of an earlier paper. In the earlier paper the authors had undertaken the addition of audio and still images with a view to augment existing text chat features for an effective online learning process. In this improved version of the application, Microsoft Agent technology in conjunction with Microsoft Text-to-Speech (TTS) engines have been used to provide synthesised voice output, which is based on an American accent at the moment but can be changed to a British accent at short notice. The next step involves the design for a Kiwi accented voice the idea for which took shape during the authors’ visit to TalkLink, situated within the Unitec campus. Here various hardware and software applications are used by TalkLink to provide synthetic voices to aid / teach their clients. These voices are either American or British accented synthetic/digitized voices in the absence of a Kiwi accented speech engine. Therefore, the authors visualised the need for a Kiwi voice engine for the in-house developed application, as well as a need to explore the use of the Kiwi Speech engine for porting on to other applications/ platforms by users such as TalkLink and the Royal New Zealand Foundation for the Blind. Since the development of a speech engine from scratch might take couple of years, the authors started exploring the possible use of available tools and resources for this purpose. During this search authors learnt that Carnegie Mellon (CM) University at Pittsburgh, USA, has a speech research centre. The researchers at CM have been working on various aspects of speech synthesis over a long time. They have developed a number of tools for developing synthetic voices. These tools are available as open source along with good documentation. This paper relates to the first of the three phases for the development of the Kiwi Voice Speech engine. The authors investigate appropriate tools for this development, from a range of tools that include Flite, festVox and Festival, developed by the research team at CM. The authors are currently analysing the tools based on the three basic parts of a TTS engine. Briefly the first part relates to text analysis for finding the basic utterance. The second part relates to linguistic analysis for pronunciation and the third part relates to appropriate waveform generation. The authors’ intenton to adopt the Festival approach is expected to reduce the cost and time for developing a Kiwi accented TTS. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.naccq.ac.nz/conference05/proceedings_04/sathu_kiwivoice.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
