A Viterbi Detector, Based on Sinusoid Modelling of Fading Mobile Radio Channels:

An Illustration of the Utility of Deterministic Models of Time-Variations in Adaptive Systems.

Lars Lindbom, Mikael Sternad and Anders Ahlén

STU Workshop on Digital Communication,
Gothenburg, Sweden, May 29-30, 1991.

In Pdf (114K)

Abstract:

Parameters of time-varying systems are often estimated by adaptive algorithms with sliding time-windows, which discount old data.

We may then face a dilemma: the use of a short data window (or, equivalently, a large adaptation gain) results in noisy estimates. With a long data window (small gain), time varying parameters are tracked with a considerable delay. A satisfactory compromise can be hard to find, if it exists at all.

To improve the accuracy, a priori information about the properties of the time-variations may be taken into account, in the form of deterministic or stochastic models,

The algorithms proposed here was motivated by the problem of tracking the channel coefficient in a D-AMPS North American digital mobile radio system. The channel can be described as a FIR filter with two complex-valued time-varying taps. These taps vary due to Rayleigh fading, with a speed determined by the speed of the mobile.

We propose to parametrize the real and imaginary parts of the two FIR coefficients by sinusoid functions with unknown frequency. The resulting algorithm consists of four separate recursive prediction error algorithms.

The parameter estimation was tested on simulated data with 10dB SNR and with each data burst beginning with a known training sequence. As compared to using recursive identification with a sliding data window, the accuracy was improved significantly, in particular in time intervals with severe fading.

Related publications:

Report from 1990, which describes the algorithms and the simulations.

Licentiate Thesis by L Lindbom, 1992, which develops the algorithm based on deterministic modelling further.

Conference paper in IEEE ICASSP'93 on using stochastic models of time-variations .

A series of four papers outlining the later development of a complete design methodology, based on stochastic models of time-varying parameters:

Design of general constant-gain adaptation algorithms.

Part II: Analysis of stability and performance, for slow and fast variations.

The Wiener LMS adaptation algorithm, a special case with low complexity.

A Case Study on IS-136 1900MHz channels.

PhD Thesis by L Lindbom 1995, presenting the general design methodology.