Exploring the relationship between degrees of self similarity and altered driving states

Combating the dangers of distracted driving is currently one of the major road safety concerns for our society. There is much being done to increase awareness on the issue and also to legislate punishment for drivers shoo get caught turning their focus away from the road, but these have not proven to fully address the issue. While cars are equipped with several other systems to keep their drivers and all nearby safe, there is a void when it comes to tools which can help keep drivers alerts, or at least to help identify the driver´s distraction states. This work seeks to unmask distracted driving by monitoring the statistical self similarity of physiological, environmental and vehicular channels of data, through the application of Detrended Fluctuation Analysis (DFA). Combining the self similarity property for several but not all the channels in the considered data, a viable predictor was generated. Implemented in large part as a Self Organizing Map (SOM) construct, the predictor confirms that self similarity contains useful information. More work is required to uncover why this is the case, as well as just how good a predictor can be generated through extending this approach.