Estimated Based On The Em Burst Communication Parameters

Burst-mode transmission is widely employed in Time-division multiple access (TDMA) mobile communication system and Ad-hoc networks. Compared with the traditional continuous-mode communication systems, the burst-mode communication is based on a short burst of data frames as the units, rather than continuous data streams. Then the receiver technologies of the burst-mode transmission, such as synchronization parameters extraction and so on, must be different from that which are used in the continuous-mode communication receivers. While its physical layer communication protocol, communications system design must take into account the capacity of the network, spectral efficiency and other factors. The research of the key technical issues is the basis for the design and implementation of the burst-mode communications systems. The parameter estimation in burst-mode communication is of vital importance to be studied and solved.In many signal processing algorithms, the EM (Expectation-Maximization) algorithm is more popular in recent years as a statistical method of calculation. It is an effective iterative calculation method to solve the"incomplete data problem"and it is root in the research of the estimation problem, especially in the Maximum Likelihood estimation problem. Because of its fast convergence characteristic, simpleness and openness and other features, the EM algorithm has been studied by many scholars and is widely used in various subject fields including information and communication in recent years.This paper deal with the parameters estimation for burst transmission, especially on the carrier frequency offset, multipath time delay and signal-to-noise ratio, which affect the performance of the communication systems deeply. On the basis of detailed analysis of EM algorithm, some new estimation algorithms of these parameters based on the EM framework are presented. The main research results include:1. A normal EM-based frequency offset estimation algorithm with the pilot and data symbols assisted jointly are proposed in order to use pilot and unknown data jointly. Different from the traditional DA estimation methods that only use pilot symbols and the NDA estimation methods that only use unknown data symbols, the new algorithm tries to seek effective and jointly uses of the known pilot and the unknown data to improve the performance of the system without affecting the spectral efficiency. Based on the EM algorithm framework, the new algorithm introduces the unknown data as the "missing data" to construct "complete data" model, through the iterative computation and maximization of the conditional a posteriori expectation of the complete data likelihood function to get the frequency offset estimation.2. To overcoming the drawbacks of the normal EM-based algorithm, an improved Recursive EM frequency offset estimation is presented. For the normal EM-based algorithm, when too many unknown data symbols are used directly, it would lead to slow convergence speed and convergence to the unhoped local maximum point, and result in the deterioration of performance. The REM introduces the data symbols with a sequential and recursive manner, which can speed up the convergence rate and get the global maximum point with considerable probability.The simulation results show that the two EM-based estimation algorithms with few pilot symbols can attain the almost same estimation accuracy as DA estimators with more pilot symbols and the estimation accuracy of it is higher than NDA estimators. Therefore, it is more flexible for burst transmission systems to employ the joint estimators to balance the estimation performance and spectral efficiency. The communication performance and spectral efficiency can be satisfied well simultaneously by choosing the number of pilot symbols in each data burst and unknown data symbols used by the joint estimator reasonably.3. On the basis of maximum likelihood estimate of the multipath time delay, the multipath time delay EM estimation algorithm is studied. The complex L-dimension optimization of multipath time delay estimation is superseded by iteration of L parallel one-dimension optimization problem of single signal parameter estimation, which reduces the complexity of the estimation.4. In order to reduce the computation complexity and to improve the convergence performance of the multipath time delay EM estimator, some effective strategies are proposed:(1) Data-preprocess in frequency domain: When the multipath time delay is not integral multiple of the sample interval Ts, it is necessary to take interpolation process to improve the accuracy of estimation. For EM algorithm, the computing complexity will be very high if the interpolation process is taken in every iterative procedure. With the DFT, the receive data is transformed into frequency domain, so the interpolation process are avoided.(2) AP-based initialization strategy: The initialization of the iteration effect not only the convergence rate, but also the convergence result. In this paper, the Alternating Projection method is employed into the initialization of multipath time delay EM estimation. The simulation results show that the AP-based initialization strategy can improve the EM algorithm convergence rate effectively.(3) SAGE-based parameter sequential updating strategy: Parameter sequential updating method can lead to more rapid convergence rate than the parallel updating. Based on the analysis of SAGE algorithm, the parameter sequential updating is implemented by changing the value of the parameterβin every iterative of the algorithm, which resulting in a rapid convergence with less computing complexity.(4) Adaptive search interval reducing strategy: The step of parameter updating will be reduced with the iterative. Based on this conclusion, an adaptive interval reducing strategy is proposed. According the parameter estimation obtained in previous iterative to determine the new search interval of the parameter in current iterative. In this way, the length of the search interval is shortened and the algorithm computing complexity is reduced.5. An EM-based SNR estimation algorithm with pilot and unknown data aided jointly is presented. With the similar strategy in the first work, be different from the DA and NDA methods, which utilizes only pilot symbol or only unknown data symbols, the estimator jointly uses pilot symbols and unknown data symbols for SNR estimation in a iteration process based on the EM framework. Compared with the DA estimators which only use the pilot symbols, the estimation performance can be improved effectively even if the length of the pilot sequence is limited because of system spectral efficiency, and the CRLB is achieved in moderate high SNR. In a large SNR domain, the joint estimation algorithm outperforms the classic blind estimation algorithm only using unknown data symbols such as MLE and M2M4 algorithms.