This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: EE 562a Homework Set 5 Due Wednesday 21 March 2007 1 (The following welldefined problems come from different sources, and the notation used may vary. Don’t let that bother you!) 1. Recursive Estimation  A Simple Kalman Filter What if we have a sequence of observations, { x ( u, i ) } ∞ i =1 , and we would like to estimate an n dimensional random vector, v ( u )? Suppose that we know the best estimate of v ( u ) based on the observations { x ( u, i ) } k i =1 and we now observe x ( u, k +1): Do we need start over and solve the new (larger dimensional) estimation problem, or can we somehow update the estimate to account for the new information provided by x ( u, k +1)? This is the subject of this problem. Let v ( u ) be an ndimensional mean zero, Gaussian random vector. Let the i th observation be the zero mean, Gaussian random variable x ( u, i ) and consider the estimation problem described above. You may assume that v ( u ) and { x ( u, i ) } ∞ i =1 are jointly Gaussian. Denote the ( k × 1) vector of observations by x k ( u ) , x ( u, k ) x ( u, k 1) . . . x ( u, 1) , and denote the unconstrained MMSE estimate of v ( u ) based on the k observations by ˆ v k ( u ) , E { v ( u )  x ( u, k ) , x ( u, k 1) . . . x ( u, 1) } = E { v ( u )  x k ( u ) } ....
View
Full
Document
This note was uploaded on 05/06/2008 for the course EE 562a taught by Professor Toddbrun during the Spring '07 term at USC.
 Spring '07
 ToddBrun

Click to edit the document details