the-BA : Virtual Voices

By Angie Newton

Advances in voice synthesis technology could provide terrorists with a wealth of new, more cunning weapons, Professor David Howard, a member of the media engineering group, warned the BA Science Festival in York.

With the current rate of progress, electronically-synthesised speech "identical to that of the human speaker" will soon become a reality, claims Professor Howard of the University of York. Within ten to fifteen years, he expects it to be possible "to replicate the speech of anybody, saying anything, based on just one or two sentences of the original voice," he added.

Technological advancements of this kind raise important social responsibility issues for scientists, where misuse could lead to impaired security of both personal information and the general public as a whole.

The new computer model aims to simulate the physical size and shape of the vocal tract, and will eventually replace the instantly-recognisable, emotionless speech heard on answer machines world-wide, which is currently generated using acoustic theory dating from the 1950s.

Often it is our vocal ticks, such as variations in the pitch and rhythm of speech, that give away our identity.

However even this computational challenge has been largely overcome, with many models already in place to imitate them accurately.

The key to creating this "organic" or "natural" speech is to mimic the changes in the muscular structure of the vocal tract that occur during speech, as revealed by MRI scanning, and to successfully join the sounds together.

To date, this has been achieved solely with vowel sounds. The next step will be to modify the model to include consonants, possibly with the help of speech therapists.

Read more news coverage at the Daily Telegraph and the Guardian.