EXPONENTIAL POWER ESTIMATION (EPE) BASED ADAPTIVE EQUALIZATION FOR STATIONARY AND TIME-VARYING CHANNELS

In this paper, an adaptive equalizer that uses normalized LMS algorith m is proposed and the normalization is being achieved by using the exp onential power estimator (EPE) (Bellanger and Evci, 1985; Bellanger, 1 987; Mathew et al., 1992). The exponential nature of the power estimat or together with the computational simplicity involved in implementing the equalizer, makes it advantageous over the constant step-size LMS and the normalized LMS algorithms which use block and recursive power estimators. The proposed EPE-based equalizer has been studied both for stationary and time-varying channels. For both these channels, the pr oposed algorithm has been found to provide better or equal performance in terms of convergence rate, convergence error and hence the bit err or over that of the aforementioned equalizers. Specifically, for chann els with large eigenvalue spread, it has a performance advantage of a factor of 2 (about) and this is valid even at SNR of about 15 dB.