According to an article in the Daily Mail, German researchers have created a library with speech patterns that will help predict whether a motorist is drunk and over the prescribed limit not by listening for vocal cues. According to research, alcohol contributes to 40 per cent of the road deaths around the world and the makers of the software argue that, if successful, it can prevent drivers from driving and prevent accidents.
Between 2007-2009 a database known as the ‘Alcohol Language Corpus’ was created by recording conversations of people with high levels of alcohol intake in a stationary car.
Presently, the database contains conversations of 162 German men and women. Computer scientists at Queens College and Columbia University have created an algorithm which will help in detecting whether or not the person driving is drunk by analysing the slurring speech patterns. Other indicators for drunken speech include stammering, stuttering and rise in the pitch of the voice.
Coupled with locks and immobilisers, this software will help in reducing road accidents by ensuring drink drivers are not able to drive the vehicle by preventing the car from getting started.
Impressive science or flawed?
Will this lead to greater convictions in the magistrates court for drink driving offences?…no, not yet anyway.
The major flaw, at this stage, appears to be that the algorithm accurate three quarters of time. According to Professor Andrew Rosenberg, of Queens College New York, this means that this will incorrectly brand one in three drivers as drunk.
More languages are aimed to be added into the database so that more researchers can benefit from it and use it for their own research.