A method for driving event detection using SAX with resource usage exploration on smartphone platform

Driver errors such as careless and aggressive driving behaviors are one of the key factors contributing to road traffic accidents. It is, therefore, essential that drivers are aware of their actions when they are in control of the wheel responsible for not only their own lives but also passengers and bystanders on the road. Driver monitoring and advanced driver assistance systems have already been utilized in fleet and logistic domain as well as built into high-end vehicles commercially available in the market. However, the majority of drivers on the road today do not have access to such systems. This paper proposes a novel methodology of driving event detection using a time series approximation algorithm known as symbolic aggregate approximation (SAX) on data collected from smartphone sensors. The use of smartphone allows the system to be easily accessible, widely available, and implemented at low cost. In addition, a resource usage exploration on a smartphone platform is conducted in order to demonstrate the flexibility of our proposed algorithm to match different smartphone specifications. Preliminary results from our experiments revealed that the precision of the proposed detection algorithm of aggressive driving events is fairly good as the precision values range from 50% to 100%. In terms of resource usage exploration, it has been found that there is a strong linear relationship between the parameter settings for data compression and the runtime of the algorithm. This is beneficial when a trade-off is required between the accuracy of the algorithm and the resource usage on the smartphone.