Research On Phase-Coded Signal Analysis And Processing For Radar & Communication Reconnaissance

Phase-coded signals are widely used in modern radar and communication systems. The phase characteristics are usually different in various phase-coded systems, so intercept receivers must resort to sophisticated signal processing techniques to achieve best interception performance according to the characteristics of the objects. Analysis and processing of several phase-coded signals widely used in radar and digital communications is intensively studied in this dissertation. The main research works are listed as follows.1. To investigate the joint detection and parameter estimation of polyphase codes derived from linear frequency modulation (LFM) used in low probability of intercept (LPI) radars, the linear characteristics in ambiguity function and time-frequency distributions are utilized and algorithms based on Radon-Ambiguity transform (RAT) and modified Radon-Wigner transform (RWT) are proposed. The Cramer-Rao Lower Bound (CRLB) of parameter estimation for P4 code is deduced as an example. Simulation results show that the two methods have similar parameter estimation performance which is close to CRLB at low SNRs.2. Parameter estimation methods of unbalanced quaternary phase shift keyed (UQPSK) signal widely used in satellite communications are investigated. The spectral correlation function of a UQPSK signal is derived. Signal parameters such as carrier frequency, chip rate, carrier phase and channel power ratio are estimated by means of the special cyclic spectrum amplitudes and phases. The estimation performance of channel power ratio is analyzed and verified by computer simulations. To explore the pseudo-noise (PN) code period estimation of UQPSK direct sequence spread spectrum (DSSS) signal with long and short period spreading codes, the method based on the second order moment of a correlation estimator is introduced. The relationship between periodic detection performance and signal parameters is derived, and simulation results show the validity of theoretical analysis.3. For estimating the parameters of the root raised cosine filter in any linear digital modulation scheme in digital communications, algorithms based on the inverse fast Fourier transform (IFFT) and least squares estimation are proposed. The estimation performance of roll-off factor is analyzed in detail. Simulations show that the proposed roll-off factor and chip rate estimators have good performance at low SNRs.4. To investigate the automatic modulation classification of Phase Shift Keying (PSK) signals widely used in TTC (Tracking, Telemetry and Control) and data links of satellite, such as BPSK, QPSK, OQPSK (Offset QPSK), and UQPSK, two classification features C₁ and C₂ are proposed based on the characteristics of the complex code sequnces. The proposed features are invariant with carrier frequency offset and carrier phase offset. A hierarchical scheme is used to classify the four signals based on C₁ and C₂. The classification performance is analyzed in detail and is verified by simulations.5. Parameter estimation of a long-code DSSS signal is studied. Denote the ratio of spreading code length P to processing gain L as K, we find that the long-code DSSS signals with K integral or not have different characteristics. An algorithm based on further processing of the autocorrelation estimator is proposed to estimate the symbol rate of a long-code DSSS signal with K not integral. Theoretical analysis and simulations show that PN code period and symbol period are hidden in the curves of the second order moment of the autocorrelation estimator. Simulation results show that the proposed algorithm is effective on symbol rate estimation at low SNRs. When K is an integer, the second order moment of the autocorrelation estimator of the long-code DSSS signal is derived, and which is shown to be different to that of a short-code DSSS signal and a long-code DSSS signal with K not integral. The maximum norm approach is introduced to estimate PN code period and timing offset. Finally, the covariance matrix is divided into several low rank matrices and eigenvalue decomposition approach is introduced to each matrix to estimate the PN sequence.