I'm confuse how to use pdf to predict next state in particle filter

How can I make predict state/equation for next state if I have state space $[C_x,C_y,H,W]$ where $C_x,C_y$ is position of centroid object and $H,W$ is height and width of object contour?

• can you precise 1) the context (estimating a bounding box around an object it seems) 2) what do you mean by PDF and 3) the kind of prior you have on the next state (like "it should be near the next position with similar size")? this would make an appropriate answer easier to write. – meduz Jan 3 '15 at 14:41
• I'm sorry for my bad english.. In some paper I read they use $$X_{k+1}=f_k(x_k,w_k)$$ how I can find the $f_k$? is it by using probability density function (pdf)? – user3794938 Jan 7 '15 at 9:23
• obviously, it means that f_k is the transition function to the next state knowing the current state and some estimate of parameters... this can be derived from prior assumptions on the kind of system you are modelling – meduz Feb 23 '15 at 12:42