Speech Systems Get a New Voice

by Christopher Jones
A voice compression system that was used by the British army is finding its way into mobile phone networks and smartcards in Europe. By Christopher Jones.

Talking to your PC or a bank machine could soon be good for more than stress relief thanks to advancements in speech recognition.

While voice recognition systems have become more popular on PCs in the last few years, the programs remain processor hogs that require training to recognize individual voices.

A British company, Domain Dynamics, claims to have developed a technology that eliminates these drawbacks. The creators say it can verify the unique characteristics of individual voices while easily fitting on a chip in a cell phone or a wallet-sized smartcard.

Domain Dynamics wants to incorporate its speech recognition technology in smartcards, security systems, diagnostic instruments, toys, and low-power, low-cost speech response and biometric verification devices.

Martin George, marketing manager for Domain Dynamics, said that the company is currently developing a 32-bit smartcard chip that would be able to identify individual voice patterns and verify the owner based on this biometric information. The company's software could also be integrated into the spare memory of a mobile phone, George said, preventing unauthorized users from making calls. The company is also trying to get the technology embedded in PC sound cards for speech recognition and security verification applications.

"You could have a PC connected to the Internet, where the person could be verified to trade in shares or banking. It has strong implications for security of electronic commerce," George said.

Domain Dynamics' core technology, called Time Encoded Signal Processing and Recognition (TESPAR), uses 29 unique symbols to represent speech sounds. Unlike conventional voice recognition systems, which use "Fourier transform" to determine how frequencies in a signal change over time, Domain Dynamics' system represents the shapes of sound waves in real time.

"The problem with looking at the frequency component is that when people speak more slowly or quickly, you need to do dynamic time alignment, which requires about 100 times more processing power than our system does," said George.

Located in Swindon, near London, the company was founded in 1990 by a military communications expert, Professor Reg King. A former director of telecommunications for the British army, King patented a technique that was used to compress speech communication over walkie-talkies used in battle.

The TESPAR technology is currently being used to gauge and monitor the accuracy and reliability of cell phone networks in England. By generating speech, the software measures how well a cell phone network is able to transmit and receive a signal -- or how well the reconstructed voice matches the original voice that was transmitted.

In addition to the business applications for its technology, George said the company would also like to see it used in toys and games. "You could take a battery-operated toy and say 'go to mummy, or turn left.'"

Newsbit furnished by:

A: NetShift Software Ltd.
A: Hughenden Yard, Marlborough, Wilts,SN8 1LT, UK
T: +44 (0)1672 511 094
F: +44 (0)1672 511 078
E: [email protected]

Thanks Anna!

Thanks Kinetic!

© 1998 Kiosks.Org.
All Rights Reserved.