Moncef Mettiji

Licentiate Thesis, Report UPTEC 92102R, 101pp, July 1992.

Abstract:

In this work, which consists of four different parts, the final aim was to detect angle modulated signals of known form in additive noise using robust detection techniques.

The introduction consists of a statement of the detection problem of known signals in additive noise. Optimal detectors in white Gaussian and non-Gaussian noise are considered. Finally, some robust detectors are presented.

The first part treats the problem of analysing the measured noise characteristics in a radio environment in the 900 MHz frequency band. Some uncertainty models for the noise, based on epsilon-contaminated Gaussian models are proposed and compared to the measured data.

The second part deals with the problem of extracting one, out of a finite number, of known signals in additive noise with partly unknown probability density function. Two types of robust detectors have been considered: the distance-based and the correlation-based detector.

In the third and last part, which is a direct application of the results from the preceding ones, we treat the detection problem of angle modulated signals corrupted by impulsive noise. Detectors based on M- and R-statistics for two types of modulation, GMSK and pi/4-shifted DQPSK, have been derived and compared to detectors optimized for Gaussian noise. A new way to compute the bit error probability by using the efficacy instead of 2/N0 is proposed.

Related publications:

Paper in IEEE Trans. IT 1997 on adaptive detection of known signals in additive noise by means of Kernel density estimators.

T. Öberg and M. Mettiji, "Robust detection in digital communications",
IEEE Transactions on Communications, vol 43, pp 1872-1876, May, 1995.