Projects / Programmes
MOBI-SYNT: Development of a speech synthesiser for embedded mobile terminals
Code |
Science |
Field |
Subfield |
2.07.07 |
Engineering sciences and technologies |
Computer science and informatics |
Intelligent systems - software |
Code |
Science |
Field |
P176 |
Natural sciences and mathematics |
Artificial intelligence |
speech technologies, speech synthesis, embedded systems, mobile terminals
Researchers (1)
no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
1. |
21321 |
PhD Aleš Mihelič |
Computer science and informatics |
Head |
2008 - 2010 |
50 |
Organisations (1)
no. |
Code |
Research organisation |
City |
Registration number |
No. of publicationsNo. of publications |
1. |
1986 |
ALPINEON R & D |
Ljubljana |
1820931 |
387 |
Abstract
Europe is one of the most advanced markets for speech technologies, which offer the possibility of automating existing services and inexpensive development of a large number of enhanced conversational applications. They also offer the opportunity for e-inclusion of persons with special needs in the modern information society, especially the elderly, blind, and visually impaired. The European Union is committed to ensuring that the necessary tools and resources are made available for all the EU languages as well as the world’s principle commercial languages, paving the way to a pervasive multilingual information society in Europe.
The major objective of MOBI-SYNT, the proposed 2-year Postdoctoral Project for the Business Sector, is to develop a high-quality Slovenian speech synthesizer prototype that can be implemented in embedded platforms of portable terminals. The main focus will be placed on finding new speech-processing procedures (esp. in rapid and efficient unit-selection) that will make it possible to automatically transform text into speech quickly, using as little memory as possible, and with high quality. We will also seek new procedures for objective quality evaluation of the final synthesized speech. The resulting embedded TTS system prototype, which will be portable across various platforms, will be designed with efficient memory consumption, and will be scalable in terms of footprint. It will be applicable in a wide range of wireless devices and multimodal terminals. Its first test users will be a group of blind and visually impaired persons who will use MOBI-SYNT with a screen reader on mobile handsets.
The extensive experience in speech technology development and embedded systems available in the Alpineon research group will help the postdoctoral candidate in carrying out the proposed project, as well as the specialized knowledge that the candidate gained while working on his doctoral dissertation. The proposed project is related to several international FP7, Eureka and COST projects, in which the research organization and the project leader are participating.
Significance for science
In the postdoc project three variants of an accelerated unit-selection methods were designed for an embedded implementation of a polyphone concatenative TTS system. I have simplified the graph-search procedures in order to accelerate unit-selection – which is the most time-consuming part of the TTS process – without significantly affecting the quality of the output speech signal. In the first two methods only the cost computation methods have been simpliefied, the found speech segment sequence is still optimal. Whereas with the third method we also sacrifice the optimality of the found sequence of speech segments in order to further speed-up the speech segment selection procedure.
The efficiency of the novel approach to rapid unit-selection has been demonstrated for Slovenian. However, the new methods can be easily applied for other languages, as well.
Significance for the country
The new efficient method for unit-selection enabled me to upgrade the Slovenian TTS engine Proteus for embedded platforms. This enabled blind and visually impaired persons to access information in a voice format over their mobile handsets, for the 1st time in Slovenian language.
Most important scientific results
Annual report
2008,
final report,
final report,
complete report on dLib.si
Most important socioeconomically and culturally relevant results
Annual report
2008,
final report,
final report,
complete report on dLib.si