- Unveiling latest developments in Automatic Speech Recognition (ASR) R&D activity
- Reaching new levels of accuracy, performance and functionality
March 31, 2004 - Cambridge, UK - Toshiba Research Europe Limited today unveiled the latest advances in its R&D programme to produce the next version of its embedded ASR engine. Enhanced Toshiba ASR middleware with new levels of accuracy and functionality will officially be available in Q3 2004 in nine languages - US and UK English, Dutch, French, German, Italian, Spanish, Japanese, Mandarin - with more planned.
Through its R&D activity, the Toshiba middleware will reach new standards for precision and faultlessness in speech recognition technology, boasting up to 95% accuracy even in very noisy environments. In addition, an enhanced version of Toshiba’s online grammar loading and editing technology will give in-car systems greater knowledge (e.g. the ability to read titles from a CD) and users more functionality (e.g. “play me ‘A Hard Day’s Night’” instead of “play me track 5”). Users will be able to define and deliver a vastly increased set of words and voice commands – from voice dialling anyone in their PDA address books to dictating emails. The new middleware will also feature an advanced grammar optimisation tool which allows application developers to create their own bespoke solutions, seamlessly and cost-effectively, using standard grammar formats and familiar development tools.
The core technology behind the ASR is being developed by Toshiba’s Corporate Research & Development Centre, based in Kawasaki, Japan. The development and optimisation of the European and US language components and the grammar optimisation tool is being conducted by the Speech Technology Group at Toshiba’s Cambridge Research Laboratory (CRL).
Toshiba ASR systems can be found in numerous in-car navigation systems, PCs and PDAs around the world. The company is able to draw on strengths from across the entire organisation to offer its customers a complete embedded solution. This comprises not only speech middleware (both ASR and Text-to-Speech) but also the requisite chip technology and processors, and worldwide customer support.
“High accuracy and rich functionality are the core criteria for a speech recognition solution. Thanks to our R&D efforts in Japan and Cambridge, the forthcoming version of the Toshiba ASR will deliver excellent performance in both fields,” says Dr. Kate Knill, Group Leader of the Speech Technology Group at CRL. “Most importantly, we’re dramatically reducing the error rate you would typically expect with speech recognition and will be able to deliver accuracy rates regularly around 95%. Combining this level of performance with the ability to offer a complete embedded solution will give Toshiba a genuine competitive edge in the market.”
For more information about Toshiba, please visit www.toshiba-europe.com/research/
Over Toshiba
Toshiba Information Systems Benelux is een dochteronderneming van de Computer Systems Division van Toshiba Europe GmbH, een leverancier van portable computers. Naast notebooks levert de organisatie mobiele en draadloze Pocket PC’s, accessoires en mobiele server oplossingen voor de zakelijke en consumentenmarkt. Voor meer informatie zie www.computers.toshiba.nl
De Computer Systems Division van Toshiba streeft bij de ontwikkeling van haar geavanceerde notebooks, PDA’s en servers naar de ultieme vrijheid van werken. Toshiba is toonaangevend fabrikant van mobiele producten uitgerust met WiFi- en Bluetooth-connectiviteit; dé wereldstandaarden voor draadloze systemen op kantoor, thuis en onderweg.
Het moederbedrijf, Toshiba Corporation ontwikkelt informatie- en communicatiesystemen, elektronische componenten en consumentenelektronica voor uiteenlopende markten. Het bedrijf heeft een positie verworven als innovator op het gebied van geavanceerde componenten, producten en systemen. Toshiba heeft wereldwijd 166.000 medewerkers en de jaaromzet bedraagt ruim 47 miljard US$ (2003).