Compressive Sensing Based Channel Estimation Method For Hybrid Satellite-terrestrial OFDM Communication Systems

Hybrid satellite-terrestrial OFDM communication systems combine the advantages of large coverage in satellite communications and high-quality transmission in terrestrial communications, thus enabling user terminals to enjoy all-day coverage everywhere of mobile communication network by choosing access networks depending on its environ-ment, satisfying the high-demanding requirements of different categories of mobile users. Radio waves will sacrifice multi-path effects and frequency-selective fading when user terminals are communicating with the satellite or terrestrial base stations, which will de-grade the signal transmission quality. Therefore, channel estimation is needed to acquire accurate channel state information (CSI), which will be used for channel equalization and coherence demodulation, and ensure the signal transmission quality. Current channel es-timation in OFDM systems is usually conducted in the frequency domain, which requires a mass of pilots periodically inserted into the subcarriers to exploit the channel fading, re-sulting in a high pilot overhead and a low level of system resources utilization.However, plenty of experiments have demonstrated that the wireless channel experiences sparsity in the time domain, the sparsity of satellite channels is stronger than that of terrestrial ones, and the sparsity of wide-band systems is more evident. With the trend of wire-less systems toward broadband, the researches of sparse channel estimation are attracting unprecedented attentions. The recently proposed theory of compressive sensing is able to reconstruct the original sparse signals by using only a few measurements, which can support the sparse channel estimation significantly.This thesis studies the compressive sensing based channel estimation methods, pro-poses a design method of the optimized pilot patterns which can exploit the channel ef-ficiently, designs a recovery algorithm which can reconstruct the CSI rapidly, and pro-poses an efficient and fast channel estimation scheme for the hybrid satellite-terrestrial OFDM communication systems by considering both the exploitation efficiency and recov-ery speed. By exploiting the inherent sparsity of wireless channels, the proposed scheme can achieve the channel estimation with the help of only a few pilot tones. Moreover, this scheme could adaptively adjust the amount of pilot tones depending on the actual chan-nel environment experienced by the user terminal, thus achieving the maximization of system spectrum dynamically, supporting the transparent radio interface for the satellite and terrestrial sub-systems, which has a certain theoretical significance and engineering application value.The research contents and their objectives can be specified as follows.Firstly, mathematical modeling for the compressive sensing based channel estima-tion.By associating the problem of channel estimation and the theory of compressive sensing, the mathematical model of compressive sensing based channel estimation has been established. Considering the complexity of wireless channels, randomly distributed pilot tones are selected to exploit the sparse channel,and the Smoothed l0-norm (SLO) al-gorithm is selected to reconstruct the CSI. Simulation experiments are conducted to verify the feasibility of this basic scheme, and to evaluate its pilot overhead as well as estimation performances, which builds theoretical and numerical basis for subsequent studies.Secondly, design method for the optimized pilot patterns.Considering the problem that randomly distributed pilot tones can only guarantee the basic effect of channel ex-ploitation and are hard to be implemented in practical systems, an estimation of distribu-tion algorithm based pilot pattern optimization method has been proposed.The method is guided by minimizing the mutual coherence of the sensing matrix, models the problem of pilot pattern optimization as a combinatorial optimization issue, and optimizes the dis-tribution of pilot tones by utilizing the estimation of distribution algorithm.Simulation results show that the pilot pattern optimized by the proposed method performs better dur-ing channel estimation.Moreover, according to different pilot numbers, this method can generate various high-quality pilot patterns, which can be utilized as the alternative ones for the adaptive selection and adjustment in different channel environments, and thus can support the high-efficient channel exploitation for the hybrid satellite-terrestrial commu-nication systems.Thirdly, convergence analyses of the SLO algorithm.Based on the fundamentals of the SLO algorithm, the essential reason why it can be used as the sparse channel recovery algorithm (i.e., it can rapidly reconstruct the approximate sparse complex-valued channel impulse response (CIR)) has been theoretically analyzed. Considering the problem of existing low-efficient iteration steps, the algorithm’s iterative process and convergence property have been theoretically analyzed with the help of the graduated non-convexity method, and reach the following important conclusions:the step length can reflect the recovered signal’s quality in real time, and its relative change can reflect the iterative efficiency in real time, which are verified by the numerical experiments. The conclusions lay a solid foundation for subsequently proposing the improved SLO algorithms, which could achieve high-efficient channel exploitation and fast CSI recovery. Lastly, efficient and fast channel estimation scheme for hybrid satellite-terrestrial communication systems.Two improved SLO algorithms are proposed and evaluated by numerical experiments based on the aforementioned SLO’s convergence property. For one thing, the Recognition SLO (R-SLO) algorithm can determine whether the current pilot can satisfy the requirement of channel exploitation autonomously, providing an impor-tant reference for determining appropriate pilot numbers.For another, the Thresholded SLO (T-SLO) algorithm can accelerate the CSI recovery process, facilitating the reduction of the signal processing delay introduced by channel estimation. Then, considering both the exploitation efficiency and recovery speed, an efficient and fast channel estimation method based on R-SLO and T-SLO is proposed for hybrid satellite-terrestrial communi-cation systems:by determining the appropriate pilot number for the current channel with the help of R-SLO, selecting the corresponding high-quality pilot patterns obtained by the estimation of distribution algorithm, the channel can be exploited high-efficiently; by de-termining the most appropriate threshold for recovering current CSI to control the iterative process of T-SLO, the CSI can be reconstructed rapidly. Simulation results show that the proposed method could learn the current channel environment by frame and adaptively utilizes the most appropriate pilot to exploit the channel, high-efficient iterative process to reconstruct the CSI, thus saving the precious frequency and time resources, improv-ing the system efficiency effectively, achieving efficient and fast channel estimation and supporting transparent interface between satellite and terrestrial sub-systems.