Hi, I have just studied Kalman filtering algorithm to reduce noise, and I have seen that a state transition matrix, the channel distortion matrix, the covariance matrix of the state equation input and the covariance matrix of the additive noise are needed... Do you know if there is a mean to use the algorithm without these parameters? In my opinion there are cases for which it's not possible to model the channel distortions or the transition between a sample x(n-1) and a sample x(n).
Kalman filtering
Started by ●May 3, 2006
Reply by ●May 3, 20062006-05-03
"thom" <soniceric@hotmail.com> wrote in message news:3ZCdnfDK9ot3X8XZnZ2dnUVZ_uqdnZ2d@giganews.com...> Hi, > > I have just studied Kalman filtering algorithm to reduce noise, and I have > seen that a state transition matrix, the channel distortion matrix, the > covariance matrix of the state equation input and the covariance matrix of > the additive noise are needed... > Do you know if there is a mean to use the algorithm without these > parameters? In my opinion there are cases for which it's not possible to > model the channel distortions or the transition between a sample x(n-1) > and a sample x(n).it all depends upon the original assumptions you are making - and if this can be predicted with a degree of certainty in the actual process
Reply by ●May 3, 20062006-05-03
The extened Kalman filter but it doesn't always work since it is non-linear. In many cases you can get the mathematical state equations before hand and make an estimate of the noise if it is stationary. If it isn't then you may as well go back to LMS methods. Tam
Reply by ●May 3, 20062006-05-03
"naebad" <minnaebad@yahoo.co.uk> wrote in message news:1146703678.319264.56880@y43g2000cwc.googlegroups.com...> The extened Kalman filter but it doesn't always work since it is > non-linear. > In many cases you can get the mathematical state equations before hand > and make an estimate of the noise if it is stationary. If it isn't then > you may as well go back to LMS methods. > > Tam >Is it impossible to predict non-staionarity in noise data - through adaptation -in a similar way to any adaptive structure? Perhaps it isnt practical?