Detection of steering events using hidden Markov models with multivariate observations

Journal article, 2016

In this article we propose a method to identify steering events, such as curves and manoeuvres for vehicles. We use a hidden Markov model with multidimensional observations, to estimate the number of events. Three signals, lateral acceleration, steering angle speed and vehicle speed, are used as observations. We demonstrate that hidden Markov models with a combination of continuous and discrete distributions for observations can be used to detect steering events. Further, the expected number of events is estimated using the transition matrix of hidden states. The results from both measured and simulated data show that the method works well and accurately estimates the number of steering events.