Variable Step Size (VSS) control for Circular Leaky Normalized Least Mean Square (CLNLMS) algorithm used in AEC

A variable step size (VSS) learning rate control algorithm based on echo leakage factor is proposed for controlling the adaptation of time-domain ldquocircular leaky normalized least-mean squarerdquo (CLNLMS) algorithm used in acoustic echo cancellation (AEC). The proposed control algorithm provides higher step size during initial convergence to speedup the convergence rate and reduces the step size slowly towards an optimal value for least misadjustment. The proposed VSS-CLNLMS and other additional components required for an acoustic echo cancellation (AEC) for hands free IPPhone running on desktop PC is designed in fixed point C and tested in real time environment. The proposed methods provides considerable improvement in steady state echo cancellation as well as stability under false or delay in nearend signal detection by double talk detector (DTD). The additional complexity requirement for the new adaptation control algorithm is marginal. The implementation of step size estimation of standard VSS-CLNMS consumes about 4.12 MHz on a Pentium Processor with Windows XP/2000/NT for 8 kHz sampling rate and filter order of 1024, where as the proposed method consumes about 4.8 MHz. Similar performance can be expected when implemented and executed on other targeted platforms.