PinFor a long time, breathalyzers have been the main device used by law enforcement to detect drunk drivers. Unfortunately, there few options available on the market to physically prevent someone from getting behind the wheel while drunk other than an Ignition Interlock Device.

This is why over the last few years, scientists have been attempting to gather enough information to accurately calculate a person’s drunken state based on a simple conversation.

Researchers from two German universities, Ludwig Maximilian University of Munich and the Institute of Legal Medicin, have been working on the Alcohol Language Corpus. These researchers gathered data to create the first publicly-available audio library of drunk speech patterns by getting 162 German men and women drunk over a two-year period, and then asking them to sit in the passenger seat of a non-moving vehicle and just talk while recording their conversations in various states of drunkenness.

The ALC system used these recordings to tell whether or not the driver was intoxicated through various speech patterns usually linked to intoxication, such as slurring, stammering, and stuttering. The system also watches out for changes in the person’s rhythm and higher pitch.

Based on this database, computer scientists from Columbia University and Queens College are currently developing an algorithm that detects if someone is drunk based on his or her speech. They are hopeful that one day, the technology may be installed in vehicles, and will be able to ask drivers a number of questions before starting the car. The drivers’ responses or speech patterns will then reflect whether or not they are drunk. If their responses show they are impaired by alcohol, then the vehicle will not start.

U.S. scientists believe that their software, like the breathalyzer-based Ignition Interlock Device, may one day be used as a preventive measure to keep a vehicle from starting if it is established that the driver is impaired. Will this piece of technology yield results anytime soon? No, probably not.

At the moment, there still remains to be a number of problems with the algorithm. The researchers acknowledge that this technology is still several years away from practical use, as a 2011 study reveals that the algorithm for drunk-detection is only 73 percent accurate in its current state – branding one in three drivers incorrectly. The database also needs to be expanded to include other languages.

A number of companies and car manufacturers have also played around with other ways of detecting drunk drivers, such as toying with the idea of sensors that measure odor, tactile, visual, and behavioral input in order to detect the state of the driver. Of course, all systems still have a long way to go in terms of integration, price, and accuracy.

The ideas, however, are indicative of a real attempt that the industry is making to find a genuine solution to drunk driving and reducing the number of alcohol-related accidents. After all, more than 40 percent of road fatalities throughout the world are believed to be alcohol-related.

Source: Inside the First Audio Library of Alcohol-Addled Speech(Which Might Just Help Stop Drunk Driving)