The paper presents a brief survey of the main approaches for solving the problem of radar waveform selection for
optimal tracking. The traditional approach to the system design has been to treat the detection and tracking subsystems as
completely separate entities. In last years some new types of radars equipped with highly agile software driven waveform
generators have appeared. This makes it possible to change the transmitted waveform at each time step for obtaining
minimal tracking errors. Thus the problem of the waveform selection can be formulated as the optimisation problem of
the following form: for given class of signals and target movement model to find the tracking filter structure and
waveform parameter sequence which result in a minimum variance estimate of the target state vector. Two main
approaches to waveform optimising have been considered in the paper: the control theoretic approach and information
theoretic approach. The relationships between sensor characteristics and tracking algorithm are discussed.
The paper presents a detection performance of the LFM radar signal using the Radon transform. Such a transform
extracts the line equation from the time-frequency image. The main problem which is considered in the paper is the
autonomous search process of the LFM signal by the intercept receiver. It is supposed that a priori information
concerning the signal duration and time delay is unknown. In this case the WVD of the signal can be splitted and
performance characteristics degrade. The computer simulations are carried out for different values of the moving
window width, search velocity and signal-to-noise ratio (SNR).
The paper describes a possible structure of the LPI radar signal classification algorithm based on using a computer
system with elements of the artificial intelligence (AI). Such an algorithm uses a combination of different signal
processing tools such as the Wigner-Ville Distribution, the Wavelet Transform and the Cyclostationary Signal Analysis.
The efficiency of these transformations with respect to different kinds of digital LPI radar signal modulation is
considered. For a final classification and parameters extraction on the base of time-frequency or bifrequency
representation the artificial intelligence methods can be used. One of the possible approaches to solving the radar signal
classification problem is to use a proposed in the paper algorithm which consists of several steps: time-frequency or
bifrequency transformations, a noise reduction procedure with using a two-dimensional filter, the RBF artificial neural
network (NN) probability density function estimator which extracts the feature vector used for the final radar signal
classification without an operator.