The system is based on the temporal and frequency analysis performed on a person in a normal and drunk state. It is necessary to calibrate the system with 3 samples at 0.0 corresponding to three time slots. As is known, the voice and speech are not the same at different times of the day. Once calibrated the system will detect the changes produced at the time of making a test, simply saying a phoneme, and a word.
In a second phase, it is intended that the system can detect vocal alterations. Simply speaking. Comparing these results with databases of voices in normal and drunken state, this detection may be possible.
The scientific bases of the project are solid, and they can be checked simply by typing in intoxicated alcohol and speech google. There are many articles and studies and many of them are directly official (.gov) who talk about how to detect whether or not there is ethyl poisoning based on the voice and its characteristics. There are parameters specifically sensitive to alcohol poisoning, and this is what we have integrated into our project. As we have already mentioned, the affections to the speech are produced from a certain level of battle and before they are not manifested.
In fact, our project is like an electronic ear that allows you to detect the variations that occur in speech, in a drunken state, in a similar way as a person would do, when it detects the affections of another drink person.