Transient and tracking performance analysis of the quantized LMS algorithm for time-varying system identification

This paper investigates the statistical behavior of the finite precision LMS adaptive filter in the identification of an unknown time-varying stochastic system. Nonlinear recursions are derived for the mean and mean-square behavior of the adaptive weights. Transient and tracking algorithm performance curves are generated from the recursions and shown to be in excellent agreement with Monte Carlo simulations. Our results demonstrate that linear models are inappropriate for analyzing the transient and the steady-state algorithm behavior. The performance curves indicate that the transient and tracking capabilities cannot be determined from perturbations about the infinite precision case. It is shown that the transient phase of the algorithm increases as the digital wordlength or the speed of variation of the unknown system decrease. Design examples illustrate how the theory can be used to select the algorithm step size and the number of bits in the quantizer.

[1]  Neil J. Bershad,et al.  New insights into the transient and steady-state behavior of the quantized LMS algorithm , 1995 .

[2]  Bernard Widrow,et al.  Adaptive Signal Processing , 1985 .

[3]  William A. Gardner.  Nonstationary learning characteristics of the LMS algorithm , 1987 .

[4]  B. Widrow,et al.  Stationary and nonstationary learning characteristics of the LMS adaptive filter , 1976, Proceedings of the IEEE.

[5]  Odile Macchi,et al.  Tracking capability of the least mean square algorithm: Application to an asynchronous echo canceller , 1987, IEEE Trans. Acoust. Speech Signal Process..

[6]  José Carlos M. Bermudez,et al.  A nonlinear analytical model for the quantized LMS algorithm-the power-of-two step size case , 1996, IEEE Trans. Signal Process..

[7]  S. T. Alexander,et al.  Transient weight misadjustment properties for the finite precision LMS algorithm , 1987, IEEE Trans. Acoust. Speech Signal Process..

[8]  O. Macchi.  Optimization of adaptive identification for time-varying filters , 1984, The 23rd IEEE Conference on Decision and Control.

[9]  José Carlos M. Bermudez,et al.  A nonlinear analytical model for the quantized LMS algorithm-the arbitrary step size case , 1996, IEEE Trans. Signal Process..

[10]  Bruno O. Shubert,et al.  Random variables and stochastic processes , 1979 .

[11]  J. Mazo,et al.  On the design of gradient algorithms for digitally implemented adaptive filters , 1973 .

[12]  Alan V. Oppenheim,et al.  Discrete-Time Signal Pro-cessing , 1989 .

[13]  Neil J. Bershad.  On the optimum data nonlinearity in LMS adaptation , 1986, IEEE Trans. Acoust. Speech Signal Process..

[14]  John M. Cioffi,et al.  Limited-precision effects in adaptive filtering , 1987 .

[15]  N. Bershad.  On weight update saturation nonlinearities in LMS adaptation , 1990, IEEE Trans. Acoust. Speech Signal Process..

[16]  C. Caraiscos,et al.  A roundoff error analysis of the LMS adaptive algorithm , 1984 .

[17]  Neil J. Bershad,et al.  On error-saturation nonlinearities in LMS adaptation , 1988, IEEE Trans. Acoust. Speech Signal Process..

[18]  D. Duttweiler.  Adaptive filter performance with nonlinearities in the correlation multiplier , 1982 .