MARKOVIAN MODELS AS APPLIED TO THE PROBLEM OF SLEEP ONSET RECOGNITION AND FATIGUE
D. V. Parkhomenko VANT. Ser.: Mat. Mod. Fiz. Proc 2008. Вып.4. С. 54-59.
As a result of advances in video engineering, all state-of-the-art sleep recognition systems rest upon driver video image processing. Such systems do not disturb the driver and are highly noise resistant. Most of such systems use the width of human eye opening and the rate of winking as the basic criteria. The paper considers a new approach to the problem of driver sleep onset recognition. This approach makes it possible to track the rate of winking and predict that the driver is likely to fall asleep. This approach uses probabilistic automatons and is widely used in speech processing. This approach is focused on the analysis of human winking and instructing the automaton in behaving (winking) like a human based on wide-range sampling. A software model implementing this approach has been developed and tested against real data. The paper discusses some simplest features of the model and results of its validation.
|