Adaptive filtering algorithms are usually employed to iteratively identify the impulse response of unknown linear systems to cater for time-varying signal statistics and to reduce arithmetic complexity. The well-known least mean square (LMS) algorithm and its input normalization variant, the normalized LMS (NLMS) algorithm, are widely used in many system identification problems because of their numerical stability and computational simplicity. However, in some applications such as acoustic signal processing and network echo cancellation, higher order adaptive filters are usually required to model the acoustic paths with long impulse responses. Partial update (PU) is an efficient technique improvement to reduce the power consumption and arithmetic and implementation complexities of the LMS and NLMS algorithms.

ZHOU Yi of Institute of Acoustics, Chinese Academy of Sciences carried out a series of studies and presented a new extension of conventional S-LMS algorithms and their convergence behaviors with Gaussian inputs and additive Gaussian or CG noises.

It can offer improved performances in adaptive system identification over their conventional LMS/NLMS-based counterparts in an impulsive noise environment. The theoretical analyses reveal the advantages of input normalization and the M-estimation in combating impulsive noise. Computer simulations on system identification and joint active noise and acoustic echo cancellations in automobiles with double-talk are conducted to verify the theoretical results and the effectiveness of the proposed algorithms.

This research result was recently published  on IEEE Transactions on Industrial Electronics (2011,58(9):4455-4469)