Application Of Time-Frequency Analysis In The Array Signal Processing

Recently, there has a flurry of activity in applying time-frequency analysis to array signal processing. Developing time-frequency spatial technology which is well suited for nonstationary signals and yields good performance superior to common subspace methods is necessary for many application area. This dissertation studies several time-frequency analysis methods and their applications in array processing. The contributions of this dissertation are summarized as follows: Spatial time-frequency distribution(STFD) is introduced and it's structure is analyzed, simulations show that STFD which act effective for FM signals impressively out-performs the common method. Spatial higher order time-frequency distribution is proposed for polynomial phase signals (PPS) and the virtual time-frequency ESPRIT is presented which have an array extension ability and a high stability. For the problem of direction finding of wide-band LFM signals , two principal approaches are derived. One based on the short WVD can resolve signals of same time-frequency distribution by using spatial smoothing. The other exploits the WVD of symmetrical array to design a novel STFD . Simulations demonstrate the efficacy and significant performance of the two proposed methods. In the scenarios of LFM signals, a new spatial time-frequency distribution is achieved by fractional Fourier transform(FRFT). Using compensation in the time domain or focusing in the frequency domain, we extend the STFD based on FRFT to direction-finding of wide-band LFM signals. Employing rectangular subarray stack and singular value decomposition, we present a algorithm to estimate 2-D DOA of LFM signals of same time-frequency distribution based on spatial time-frequency distribution constructed by STFT. Evolutionary spectrum is described and a new estimation method termed DFT-EP is formulated . Spatial evolutionary spectrum distribution is derived and used to estimate direction of signals which have constant or time-varying amplitude. Section short EP is advised to realize DOA estimation of wide-band FM signals,while DFT-EP can also be used to resolve the same problem in frequency domain. It is shown the latter have better performance with respect to noise and model error. A new closed-form algorithm to obtain parameter estimation of LFM signals is developed on section-dechirp direct dechirp and Beamspace-ESPRIT. It's attendant reduction in computation complexity make it applicable in the engineering implementation. Joint estimation of DOA and polarization parameters is considered. Spatial polar time-frequency distribution(SPTFD) is first established by employing a array of electromagnetic sensors and a algorithm to obtain joint estimation of 2-D doa and polarization parameters is presented. The proposed scheme based on electromagnetic sensors can resolve cyclically ambiguous estimation from sparse array. From the DOA information embedded in the intrinsic directionality of electromagnetic sensors, a closed form estimation of DOA can be achieved even for arbitrary arrayc sensors. In most high-resolution direction-finding methods, such as subspace methods or time-frequency spatial approaches, additive noise is assumed to be Gaussian. If the noise model have bias from Gaussian, the performance of aforementioned methods degrade sharply. We present a new method for bearing estimation of LFM signals embedded in SaS noise based on FRFT and Fractional lower order moments(FLOM) . Simulation results verify the efficacy of the method in the scenarios of SaS noise .