A Stochastic Total Least Squares Solution of Adaptive Filtering Problem

Abstract EN

An efficient and computationally linear algorithm is derived for total leastsquares solution of adaptive filtering problem, when both input and output signalsare contaminated by noise.

The proposed total least mean squares (TLMS) algorithmis designed by recursively computing an optimal solution of adaptive TLS problem byminimizing instantaneous value of weighted cost function.

Convergence analysis of thealgorithm is given to show the global convergence of the proposed algorithm, provided thatthe stepsize parameter is appropriately chosen.

The TLMS algorithm is computationallysimpler than the other TLS algorithms and demonstrates a better performance as comparedwith the least mean square (LMS) and normalized least mean square (NLMS) algorithms.

Itprovides minimum mean square deviation by exhibiting better convergence in misalignmentfor unknown system identification under noisy inputs.