RESEARCH IN DETECTION THEORY OF THE STOCHASTIC SIGNAL IN NOISE
Main researcher
: UGRINOVIĆ, KOSTA (112456) Assistants
MILIČIĆ, ANDRINA (182954)
Type of research: applied Duration from: 06/01/92. to 05/31/95. Papers on project (total): 6
Papers on project quoted in Current Contents: 1
Institution name: Fakultet prirodoslovno matematičkih znanosti i odgojnih područja, Split (177) Department/Institute: Department of mathematics and informatics Address: Nikole Tesle 12 City: 21000 - Split, Croatia
Communication
Phone: 385 (0)21 587133
Fax: 385 (0)21 362431
E-mail: ugrin@mapmf.pmfst.hr
Phone: 385 (0)21 587009
Summary: Detection of the sophisticated stochastic signal in noise
is considered. The signal is supposed to be the sum of two kinds of
components: wideband and narrowband ones. Due to this assumption it was
defined the mathematical model of the signal power spectral density. So
defined the model became the basis of the new synthesis method regarding to
the sophisticated signal detection optimum structure. In order to
synthesize the optimum structure both the likelihood ratio statistical test
and the Neyman-Pearson criterion are used. To evaluate the optimum
structure performance the ROC (Receiver Operating Characteristics) diagram
are investigated and compared with the suboptimum structures ROC diagrams.
Keywords: detection theory, stochastic signal, statistical test, likelihood ratio
Research goals: Taking into consideration all in detection theory
about stochastic signals in noise it was intended to derive a general
algorithm for the detection of sophisticated stochastic signal in noise.
The stochasic signal was defined as a sum of two or more statistical
independent parts, and it was supposed to be ergodic Gaussian process of
long time duration. For this assumption the discrete signal samples in the
frequency domain are statistical independent, and on this basis the
optimum detection algorithm was researched. In the research work the
likelihood ratio statistical test and the Neyman-Pearson criterion were
used. The research result go to show that the indirect measure of the
detection probability depends on the number of discrete frequency samples
and on the power spectral densities input ratio of additive stochastic
signal components and noise. Other information about the project.