Methods to update the coefficients of lattice filter and its application to system identification

摘要

This paper proposes a method to update the reflection coefficients of lattice filter by using adaptive algorithms for finite impulse response (FIR) filter. The reflection coefficients are generally estimated as the cross-correlation coefficient between the forward and backward prediction errors in each stage of lattice filter. The estimation therefore requires two divisions in each stage totally double the number of steps. Unfortunately, the division requires far higher 1tlocessing cost than that of multiplication. This high processing cost becomes an obstacle to the application of lattice filter to practical systems. The method, proposed in this paper, can decrease the number of divisions to one provided that using the normalized least mean square (NLMS) algorithm. On the other hand, conventional method to update the reflection coefficients cannot be used as the adaptive algorithm for system identification. However, the method enables to apply the lattice filter to system identification. The stability of lattice filter can be easily discriminated. The method simplifies structuring the inverse filter.