PARIS, November 13 (IslamOnline & News Agencies) - Experts said Wednesday, November 13, an audio tape purportedly made recently by Osama bin Laden would be technically tough to authenticate 100 percent because of the recording's poor sound quality.

Only if the recording is top quality can "the best voice recognition systems [be] 99-percent accurate", said Herve Boulard, a professor at the Federal Polytechnic School in Lausanne, Switzerland, who is also head of a research institute, IDIAP, that does work for the Pentagon and other defense organizations.

In the case of the latest tape, the "result is likely to be 70 percent, which is low," Agence France-Presse (AFP) quoted him as saying.

In an audiotape attributed to him and broadcast by Al-Jazeera television late Tuesday, November 12, the supposed Bin Laden blasted U.S. President George W. Bush as the "pharaoh of the century" and his key allies as "murderers."

The expressions used, as well as the locution and tone of voice, have left little doubt as to the identity of the speaker, according to preliminary analyses by U.S. and Japanese voice experts.

Japanese specialist, Matsumi Suzuki, said the tape had been recorded in a room and the voice was hoarse, suggesting that if this were indeed Bin Laden, his health had deteriorated.

Patrick Macron, chairman of Arts Techniques, a small company in the Internet security business based in the northwestern French city of Rouen, said voice recognition software worked by taking a sample recording of that person.

Key words are sought out and analyzed for pitch, amplitude and inflection, which are then turned into mathematical values.

Those factors go into a model called an algorithm - a vocal "signature" of that person.

This algorithm is then compared against a similar model drawn from the new material, based on the same words.

"A simple example is in mobile phones which are able to recognize a name when it is spoken, and then dial that person's number," he said.

More sophisticated hands-free devices, such as microsurgery devices or weapons-firing systems in warplanes, use the same principle.

But the system is not infallible.

If someone has a cold or a croaky voice, that can distort the algorithm, making it difficult to see if there is a match.

Voice analysis does not just entail a comparison of algorithms, he said.

Other powerful software tools are brought in to see whether the recording is a montage of sound bites or has been recorded in one session or more, which could point to a hoax or throw up other details.

A one-off recording session, for instance, adds weight to the authenticity.