LTJ adaptive filter with fixed reflection coefficients for speech applications

This paper presents a lattice transversal joint (LTJ) adaptive filter whose reflection coefficients are fixed. Reflection coefficients are given by the statistics of speech signal and the proposed order of the lattice predictor is one. The additional computational complexity over the transversal adaptive filter is only 1 addition and 2 multiplications per sample. Simulation results verify that the proposed LTJ adaptive filter achieves fast convergence with negligible additional complexity compared to the transversal adaptive filter.