ADAPTIVE PARAMETER ESTIMATION USING PARALLEL KALMAN FILTERING FOR SPREAD SPECTRUM CODE AND DOPPLER T

Abstract

We present a new digital direct-sequence (DS) re- ceiver with joint estimation of code delay, multipath gains and Doppler shift. A parameter estimator consisting of a parallel bank of extended Kalman filters (EKF's) extracts estimates of the timing, T and the multipath coefficients, fl distorting the received signal. A detected estimate of the Doppler shift, UT distorting the received signal is also provided by the estimator. We compute the bit error rate that results when a RAKE matched filter uses the estimated parameters to detect the DPSK encoded binary data in the received signal. The bit-error rate (BER) is evaluated, and successful performance of the proposed receiver in the presence of Doppler shift distortion is observed in many cases. We demonstrate that the receiver can operate when the multipath coefficients vary in time (Doppler spread). N this paper we apply a RAKE receiver to the problem of I spread-spectrum communication in multipath. The reader is referred to (l) for an introduction and survey of spread spectrum RAKE type receivers in multipath. For the theory of extended Kalman filtering and adaptive parameter estimation the reader is referred to 131. In order for a spread-spectrum receiver to demodulate data, the timing offset between the receiver generated reference signal and the transmitted signal must be accurately tracked. Any Doppler shift distortion present on the signal must also be taken into account. Here we will assume a pre-acquired coarse timing estimate. The remaining problem is thus to acquire coarse estimates of the Doppler shift, and to track the code delay, timing and complex-valued multipath coefficients. The problem of tracking timing and multipath coefficients jointly was first considered in 141 and solved using an EKF algorithm. The EKF in (5) tracked Doppler by appending the Doppler parameter to the state vector, and was thus subject to filter divergence. In light of the divergence problems encountered in (5), we seek an alternative estimator which can both acquire and track Doppler shift. We will apply the general technique of partitioning de- scribed in (6), and (3) to estimate w,. The motivation for this approach is the poor observability of the Doppler shift param- eter, which can lead to filter divergence in an EKF algorithm. In summary, a bank of Kalman filters; each conditioned on a particular value, v,, i of the unknown Doppler velocity, v, can