December 5, 2014 at 15:30The Germans created a library of drunk audio recordings
Epigraph:
- Drank?
- I didn’t drink it!
- Tell Gibraltar.
- Drank.
The topic is a bit curious, but it seems to me that these are the ones that have recently been missing on Habré. Especially on Friday.
So, German scientists from two universities in Munich have created an audio recording database, which recorded the speech of 162 people intoxicated. Data was collected from 2007 to 2009 and now a language corpus is created on the basis of the database ( an article on corps on the Wiki) Alcohol Language Corpus (ALC).
The case is supposed to be used for machine learning, so that automation systems (for example, in a car) automatically recognize the state of the owner and block control. Thus, it is supposed to deal with the unscrupulous human factor.
Prior to this, several automobile concerns worked on the detection of drunken behavior. But then it was either about sensors that capture alcohol vapor (and sometimes falsely triggered in the presence of a drunken passenger), sweat analysis on the shift knob, or cameras that gaze intently into the driver's face. But, apparently, no one has thought of analyzing speech yet.
The cost of a set of records is about $ 1200 (an impressive amount for 162 structured speeches of German alcoholics :)
By the way, here is an examplefrom the base.
The news is over, you can put a minus for brevity.
- Drank?
- I didn’t drink it!
- Tell Gibraltar.
- Drank.
The topic is a bit curious, but it seems to me that these are the ones that have recently been missing on Habré. Especially on Friday.
So, German scientists from two universities in Munich have created an audio recording database, which recorded the speech of 162 people intoxicated. Data was collected from 2007 to 2009 and now a language corpus is created on the basis of the database ( an article on corps on the Wiki) Alcohol Language Corpus (ALC).
The case is supposed to be used for machine learning, so that automation systems (for example, in a car) automatically recognize the state of the owner and block control. Thus, it is supposed to deal with the unscrupulous human factor.
Prior to this, several automobile concerns worked on the detection of drunken behavior. But then it was either about sensors that capture alcohol vapor (and sometimes falsely triggered in the presence of a drunken passenger), sweat analysis on the shift knob, or cameras that gaze intently into the driver's face. But, apparently, no one has thought of analyzing speech yet.
The cost of a set of records is about $ 1200 (an impressive amount for 162 structured speeches of German alcoholics :)
By the way, here is an examplefrom the base.
The news is over, you can put a minus for brevity.