文档介绍：DSP Journal, Volume 9, Issue 1, June, 2009
An FPGA-Based Design of Fixed-Point Kalman Filter
      
Ruchi Pasricha1, Sanjay Sharma2
1Assistant Professor, CEC, Landran
2Assistant Professor, Thapar University, Patiala
Abstract representation (or equivalently in the Kalman filter)
In this paper we study scaling rules and round-off may be associated either with the Levinson recursions
noise variances in a fixed-point implementation of the for factoring the inverse of the correlation matrix for
Kalman filter for an ARMA time series observed the time series or with the LeRoux-Gueguen
noise free. The kalman filter is realized in a fast form recursions for factoring the correlation matrix itself.
that uses the so-called fast Kalman gain algorithm. The latter association leads to a fixed-point algorithm
The algorithm for the gain is fixed point. Scaling puting Kalman gains. This so-called fast
rules and expressions for rounding error variances are algorithm produces a fast Kalman filter.
derived the numerical results show that the fixed- In this paper, we present results from a study of fast
point realization performs very close to the floating Kalman predictors, implemented in floating point and
point realization for relatively low-order ARMA time in fixed-point arithmetic, for autoregressive moving
series that are not too narrow band. The floati