SPEECH VERIFICATION NOW READY FOR BUSINESS APPLICATION

Speech verification (SV) technology - using individuals’ voices to verify a speaker’s identity - is now ready for deployment in well designed, low risk applications.

This is the conclusion of a European Union research project called CAVE (Caller Verification in Banking and Telecommunications) in which Vocalis, the Cambridge based speech recognition and telephony company, has played a central role.

The CAVE project has just been completed, culminating in tests of two SV systems, a telephone banking system in Zurich, Switzerland, and a telecommunications calling card application in the Hague, Netherlands. CAVE was intended to demonstrate that analysing a caller’s speech can identify the caller and combat fraud in areas such as banking and telecommunications. Technology provided by Vocalis, a world leader in speech recognition for telephony, has therefore been key to the successful performance of the CAVE systems.

"The success of CAVE, and of SV’s overall development, depends critically on the quality of the speech technology used," comments Trevor Thomas, Chief Scientist at Vocalis. "CAVE has demonstrated the viability of using speech technology in practical SV applications, and is a further instance of our long history of involvement in industrial research projects involving voice processing."

Several other leading European organisations took part in CAVE including the Union Bank of Switzerland and the Netherlands’ PTT Telecom, which staged the tests. The results of CAVE have been so encouraging that the participating companies have decided to continue the research initiative with a second phase of the project. This project is called PICASSO (PIoneering Caller Authentication for Secure Service Operation), and Vocalis will play a further major role here too.

CAVE has achieved important results, according to the final report which has just been published. CAVE resulted in "very substantial improvements of the base line performance of the SV technology," the report says. The report states that existing SV algorithms have been enhanced, especially in cases where enrolment data is limited and that equal error rates have been reduced.

Other important advances have been made in the areas of human factors, such as dialogue design, and in analysing user requirements. For instance, CAVE has shown that in potential SV applications, service managers regard ease of use and robustness as important as security.

"The EU funding of this project underscores the critical importance of speech technology to businesses in Europe, now and in the future. " Trevor Thomas explains "At Vocalis, our R&D department is involved in many wide ranging projects to ensure we stay at the leading edge of next generation speech technologies, a market which has exciting growth potential into the next century."

Vocalis is currently involved in a series of important speech-related EU projects including:

• REWARD (Real World Applications of Robust Dialogue), which focuses on the development and design of speech-based dialogue systems to be used in automatic telephone systems
• SPEECHDAT (SPEECH DATabases for creation of voice driven tele-services), an initiative concentrating on producing speech databases
• DISC (spoken language DIalogue Systems and Components), a model of best practice procedures and methods for spoken language dialogue systems covering speech recognition, human factors and systems integration to establish dialogue engineering standards

"The unifying aspect of each of these EU initiatives involving Vocalis is the calibre of partners involved, including leading companies like Taylor Nelson, Telia Respons, Siemens, BT, and Daimler-Benz," said Trevor Thomas. "This combined research, sharing resources and information between major players in the voice-driven technology field, means we can make rapid progress and also form a more rounded picture of the research areas involved."

- ENDS -