It is considered that about 40% of traffic accidents with a fatal outcome is caused by drunk driving. This prompted German scientists to explore the way to prevent these accidents. They have developed a software which should check whether the driver is drunk and whether it is capable of driving a motor vehicle. To assess this, this software uses voice patterns.
Between 2007 and 2009, German researchers have created a library which contains conversations of 162 German men and women who had an increased level of alcohol in the blood. This database is known as Alcohol Language Corpus. Based on these data, computer scientists from Queens College and Columbia University have created an algorithm that needs to check, based on the peculiarities of speech, whether the person is drunk and whether it is capable of vehicle handling. These characteristics and patterns of speech include stuttering, changes in tone of voice, smearing speech. Based on this algorithm, it has been developed a software which is connected to the car and the main controls. If the software detects that the driver is drunk, it blocks the command and disables the start of the vehicle.
This in itself is not a bad idea, but for its implementation is needed more testing and researching. One of the major flaws of this software is its reliability, which is 3/4. In other words, this means that one in four people will be wrongly estimated. This reliability must be higher in order to the product be useful. In addition, in the software should be incorporated much more samples, and the different languages. Thus, the algorithm will be better and more accurate, and the software will be more accurate in the estimation. Until then, this may be just an interesting idea and a good attempt of solving this global problem.
For more useful articles like these check out the Criminal Barristers London website.