Performance analysis of two EM-based measurement bias estimation processes for tracking systems

Measurement Bias Estimation in the Problem of Target Tracking

“…In [15], the measurement bias is incorporated into two different state-space models, for which two EM-based estimation processes are, respectively, given based on the results of EKF and extended RTS smoother. However, they are only suitable for strictly stationary bias.…”

“…Therefore, the EM estimation should be adapted to cope with the biases having asynchronous stationary time intervals. More precisely, the stationary time interval for the EM estimation, i.e., [ , , , , , ] in (15), is different for different values of i and j, which is determined by bias switch detection. In Section 5, we give an example of asynchronous execution of the EM estimation based on Figure 4.…”

“…In contrast, the bias is treated as a random variable with limited fluctuation in different filtering frameworks [12][13][14]. In our previous work [15], the TOA measurement is employed such that a nonlinear state-space model is used to formulate the tracking problem [16]. The measurement bias is assumed to be strictly stationary Gaussian distributed and incorporated into two different state-space models.…”

“…In this paper, we extend the work in [15] and consider a more complicated case where the bias is assumed to be a random variable with a locally stationary Gaussian distribution. For this, the random variable follows a Gaussian distribution in a time interval, but in different time intervals the Gaussian distribution has different forms, i.e., with different means and variances.…”

“…In order to estimate the measurement bias which follows a locally stationary Gaussian distribution using the EM algorithm, we have to cope with two issues in addition to the work in [15]. First, the stationary time intervals in which the bias follows different Gaussian distributions should be divided appropriately.…”