Tuesday, March 11, 2014

PRIVACY: How to fool voice recognition programs (with video)

Martin Gardiner in Improbable Research
Voice_ChangerGiven the current interest in the implications of automated computer analysis of voice recordings, say for example telephone calls, it’s perhaps not surprising that some might have thought about disguising their voice to (try to) avoid recognition and/or obfuscate the content. 

But those who work professionally in the voice-recognition field have been investigating such nefarious methodology for more than a decade. In a now-classic paper on the subject published as long ago as 2000, authors professor Robert D. Rodman and Michael S. Powell of the Department of Computer Science at North Carolina State University, US, listed more than 20 viable types of voice disguise techniques. 

They include, for example:



● Putting on a foreign accent
● Deliberately lisping
● Tongue holding (whilst speaking)
● and so-called ‘Pipe-smoker’ speech (talking whilst biting a hard object)

A video example of which is available here. Compare and contrast various phrases with and without pipe.


See: ‘Computer Recognition of Speakers Who Disguise Their Voice’ in : Proceedings of the International Conference on Signal Processing Applications & Technology 2000 (ICSPAT2000), October, 2000.

Note: The photo above shows the 12018 ‘voice changer’ available from Berwick Industrial Co. Ltd. Hong Kong.