Multi-Parameter Estimation Of Radar Signals Based On Time-Frequency Distribution

Time-frequency distributions were introduced as a means of representing signals whose frequency content is varying with time, and for which both time domain representation and frequency representation are inadequate to describe the signal appropriately. This dissertation is intended to incorporate time-frequency analysis into array signal processing for the estimation of radar signal parameters. Compared with traditional solutions that try to analyze signals received by single sensor, in the situation of radar signal estimation the source array manifolds make it possible to improve time-frequency representation through time sequences obtained through multi-channel. The estimation of direction of arrival, extraction of time-frequency characters, algorithms for efficient time-frequency computation and the methods to estimate instantaneous frequency are discussed.1. However the Cohen class time-frequency distributions can provide better concentration, they are suffered from the cross-terms and the noise in the time-frequency plane. Different from traditional practice that tries to deal with this problem through samples obtained from single channel, the approach to improve time-frequency distribution using averaged cross Wigner-Ville distributions between the sensors at symmetrical places is investigated, which tries to make full use of the advantage of sensor array. Compared with the array averaged Wigner distribution, it can provide better performance and compensate the time delays among the signals impinging on different antennas. Since the noise arrives at each sensor is uncorrelated. The average of cross Wigner distributions can effectively improve the time-frequency representation.2. To compute time-frequency distribution is not the aim but just a preparation for further estimation. So we consider the post process of TFD. Unlike the methods such as computational kernel design, the time-frequency representation can be considered as an image where the pixels correspond to the time-frequency points and their intensity to the magnitude of the transform. From the viewpoint of image processing, the auto-terms of the signal can be defined as objects in the image while all the interfering cross terms and the noise form the background in the image. Therefore it is natural for the introduction of image processing technique to time-frequency representation area, and the attempt to promote the time-frequency representation is equal to extract theseobjects from the resultant image with background suppression.3. Exponentially forgetting transform (EFT) is discussed in detail in this paper, as an efficient recursive algorithm .to compute time-frequency distribution. In this approach, instead of excluding the old samples, their importance is diminished by using a special computational window, thus recursive operations can be introduced which incorporate knowledge of the time-frequency distribution from previous data blocks into the current estimate and greatly increase the computation efficiency. Under its analysis, we propose the modified version, which has the simplest recursive computation structure. The performance of EFT is investigated, which indicates that time-frequency distributions using single-sided window could not accurately trace the variation of the instantaneous frequency, since their performance depend only on the history of the signal, so they are biased frequency estimator. The excursion of instantaneous frequency representation will be resulted when EFT is applied to signals with rapidly varying frequency contents.4. The double-sided exponentially forgetting transform (DSEFT) is also presented which offers better time-frequency resolution. Compared with other forms of time-frequency representation, DSEFT is highly computational efficient. Analysis shows that DSEFT is superior to traditional exponential forgetting transform as an instantaneous frequency estimator. The bias and the mean square deviation of the estimator are much less than that of the previous one. An...