13/09/2012 - Text to Speech expert CereProc unveils findings to peers at InterSpeech 2012

A senior team from text to speech (TTS) company CereProc will today present their paper Proper Name Splicing in Computer Games with TTS to delegates at Interspeech 2012, a global leading conference on speech technology, taking place in Portland, Oregon.

Informing a prestigious audience of speech technology experts, scientists and industry engineers attending the conference, experts from CereProc will outline their strategy for producing a 'smart splicing' synthesiser for proper names, evaluating two different systems created for this application, and finally discussing the implications of its findings with regards to a wider application area within computer gaming.

Keen to unveil their findings at the upcoming global technology conference, Dr. Matthew Aylett, CTO at CereProc, said: "It's great to be attending this year's Interspeech. The conference is always a fantastic event for all delegates and speakers. Providing us with the opportunity to meet with key influencers and discuss the key issues affecting our industry today, to showcase our latest findings and inform of future technological developments on a global stage, and to attend the sessions delivered by peers from the international speech technology community."

Identifying the need to address the complex challenges presented as users of video games more and more frequently elect to play using a mobile device, the CereProc paper explores the growing requirements for smart speech splicing, and for using speech synthesis in constrained - but not closed - domains. CereProc lay out an innovative approach - using an innovative automatically trained targeting system based on parametric synthesiser HTS - that can be adapted to meet such demanding and intricate requirements.

"We're becoming increasingly aware that customers expect speech synthesis to solve a range of problems that differ from the classic application of text to speech technology. With expectations demanding speech output to be spoken correctly, when given some unknown text and limitations, we believe it vital to explore and develop solutions in these early stages. The video games market holds real opportunity for those involved in speech synthesis, and it's appropriate for us to apply our approach within this context," Aylett added.

Shown to perform well for limited mounts of data, and for data collected under imperfect conditions - in comparison to a concatenative monophone system and a baseline concatenative diphone system - the methodology implemented by CereProc - and explained during their session at Interspeech 2012 - utilises the adaptability of the automatically trained targeting system based - parametric synthesiser HTS - to carry out analyses and explore solutions that allow advance speech synthesis systems to operate as speech editing systems.

Interspeech 2012 takes place in Portland, Oregon between 9 - 13 September 2012.