From Wikipedia, the free encyclopedia
Jump to: navigation, search
Developer(s) CereProc  Scotland
Initial release 2006
Written in C
Operating system Cross-platform
Available in English / French / Castilian Spanish / German / Italian / Catalan / more coming soon
Type Text-To-Speech
License Commercial

CereVoice is a multi-lingual speech synthesis system developed at Scottish IT company CereProc. It offers a full text to speech system with various APIs.

Inside CereVoice[edit]

IVONA text-to-speech system was described at Blizzard Challenge 2006. [1] and Blizzard Challenge 2007 (special version for Blizzard Challenge). [2] It is composed of two parts: a front-end and a back-end. The front-end has two major tasks. First, it converts raw text containing symbols like numbers and abbreviations into the equivalent of written-out words. This process is often called text normalization, pre-processing, or tokenization. The front-end then assigns phonetic transcriptions to each word, and divides and marks the text into prosodic units, like phrases, clauses, and sentences. Phonetic transcriptions and prosody information together make up the symbolic linguistic representation that is output by the front-end. The back-end—often referred to as the synthesizer—then converts the symbolic linguistic representation into sound.

Unit selection synthesis[edit]

IVONA uses Unit Selection with Limited Time-scale Modification (USLTM) described in their Blizzard Challenge 2006 paper[1]. Unit selection synthesis uses large databases of recorded speech. During database creation, each recorded utterance is segmented into some or all of the following: individual phones, syllables, morphemes, words, phrases, and sentences. The division into segments is done using a specially modified speech recognizer.[3] An index of the units in the speech database is then created based on the segmentation and acoustic parameters like the fundamental frequency (pitch), duration, position in the syllable, and neighboring phones. At runtime, the desired target utterance is created by determining the best chain of candidate units from the database (unit selection).

Unit selection provides the greatest naturalness, because it applies digital signal processing (DSP) to the recorded speech only at concatenation points. DSP often makes recorded speech sound less natural.

Generated speech quality[edit]

IVONA Text To Speech System received the highest Mean Opinion Score (MOS) at the scientific contest Blizzard Challenge 2007 in Bonn, Germany. The sentences read out by IVONA were evaluated by experts, a group of British and American students and volunteers recruited via the Internet. Average mean opinion score for IVONA was the highest (3.9 points) from all speech synthesizers. A real person’s recording scored 4.7.[4]

IVONA was also evaluated at Blizzard Challenge 2006 in Pittsburgh, USA and received best Mean Opinion Score (MOS) provided by Speech Experts and Undergraduates for full database results.[5]

Voices and languages[edit]

IVONA currently speaks seven different languages with nineteen voices. [6]
American English: Kimberly, Kendra, Jennifer, Joey, Eric and Chipmunk Skippy
American Spanish: Penélope, Miguel
British English: Emma, Amy and Brian
German: Marlene, Hans
French: Céline, Mathieu
Castilian Spanish: Conchita, Enrique
Polish: Maja, Ewa Jacek and Jan
Romanian: Carmen

System compatibility[edit]

IVONA is compatible with Windows and Unix based systems.

See also[edit]


External links[edit]

Category:Computer accessibility Category:Speech synthesis Category:Multimedia software