The robust adaptive filtering is researched when the input and output signal are both corrupted by noise in this paper.On the basis of minimizing total mean square error (TMSE) and the steepest descent principle

by modifying the gradient of TMSE

a robust total least mean square (RTLMS) adaptive filter algorithm is proposed.The performance analysis compared with other algorithm demonstrates the proposed algorithm can efficiently reduce the sensitivity of the weight adjusting-tap to noise.The simulation results have also shown that the robust anti-noise performance and the stable convergence precision of the proposed algorithm are remarkably higher than other congener algorithms.And it can still keep nice convergence when a larger learning factor is used and the noise is strong.