Stereophonic Acoustic Echo Cancellation Using NLMS with Orthogonal Correction Factors

A stereophonic echo canceler is proposed based on the Normalized LMS algorithm with orthogonal correction factors (NLMS-OCF). The echo canceler is modeled using a two-input single-output finite-impulse-response (FIR) structure. NLMS-OCF updates the echo canceler coefficients based on multiple input vectors, while NLMS adapts the coefficients based on a single input vector. The proposed algorithm is simulated in MATLAB with two different types of source signals in the far-end room, namely white noise and USASI noise. Simulation results indicate that the NLMS-OCF algorithm produces faster convergence than the well-known stereo projection algorithm, in terms of faster improvement in echo-return loss as well as faster reduction in impulse response misalignment. 1