Online edition of India's National Newspaper
Saturday, April 15, 2000

Front Page | National | International | Regional | Opinion | Business | Sport | Entertainment | Miscellaneous | Features | Classifieds | Employment | Index | Home

National | Previous | Next

Need for foolproof speech identification technology

By Our Staff Reporter

NEW DELHI, APRIL 14. Even as police are going ahead with investigations in the match-fixing scandal on the basis of recorded conversations between the sacked South African cricket captain, Hansie Cronje, and bookie Sanjeev Chawla, a foolproof speech identification technology is not available in the country.

Though a speech identification laboratory was being set up at the Central Forensic Science Laboratory in Chandigarh, a case of this nature was yet to be brought before the laboratory. Scientists feel that it is not possible to ascertain with surety whether the voice on the mobile phone was that of Cronje.

Dr. S.H. Aggarwal, a speech scientist with the Centre of Electronic Engineering and Research Institute here and an advisor to the CFSL project, said despite a lot of research on the subject throughout the world, speech identification of a recorded voice should not be permissible as evidence in the court of law.

The ongoing project at Chandigarh, Dr. Aggarwal said, had achieved very little despite many efforts in the experimental stage. Drawing a parallel with the lie detector test, he added that such evidences could not be made the basis to punish an accused.

While not denying the incapability of the CFSL to carry out the tests, its Deputy Director, Dr. C.N. Bhattacharya, said proper voice samples of Cronje were needed and before that the quality of recording done by police had to be examined carefully for any test to be carried out.

The test would apparently be more difficult in this case as both- Hansie and Sanjeev- were conversing on mobile phones. The voice, according to Dr. Bhattacharya, could be affected by external disturbances which would have to be filtered repeatedly without damaging the recording. ``This was a very tedious job that could take a long time.''

The five most important factors matched during voice identification were pitch, nasality, bandwidth of the format, higher harmonics and resonance. These have to necessarily match during voice identification. While recording the voice of Cronje, according to Dr. Aggarwal, conditions similar to the ones during the conversation between him and the bookie need to be created for a more accurate result. In fact, he said, the medium of mobile phone should also be retained.

However, Dr. Aggarwal admitted that any identification carried out through voice sampling was ``tricky'' and could best serve as supportive evidence in the case. The reason was the variability of human voice which could change due to factors such as environment, mood, noise conditions among others.

Send this article to Friends by E-Mail


Section  : National
Previous : Call for more investment in public health sector
Next     : Names of 7 more bookies surface

Front Page | National | International | Regional | Opinion | Business | Sport | Entertainment | Miscellaneous | Features | Classifieds | Employment | Index | Home

Copyright © 2000 The Hindu

Republication or redissemination of the contents of this screen are expressly prohibited without the written consent of The Hindu