Research On Compressed Sensing Channel Estimation Techniques In OFDM Systems

Accurate channel estimation is a necessary condition for the data demodulation at the receiver in OFDM systems. Traditional channel estimation methods can be divided into two groups:pilot-aided channel estimation method and blind channel estimation method. However, both of theses two methods suffer from some drawbacks in practice. The first method needs to use a lot of resource for the transmission and reception of pilot symbols, which reduces the spectral efficiency of the system, while the complexity of the second method is always high, thus the implementation is difficult due to the limitation of the system hardware. How to achieve a good tradeoff between the estimation performance and the complexity is becoming an important problem in the area of channel estimation. In order to overcome this problem, in this dissertation, new channel estimation schemes based on compressed sensing are studied, and the main contents are described as follows.Chapter 1 introduces the research backgrounds, main research contents and structures of the dissertation.Chapter 2 discusses the main principle, related theratical conclusions and some recovery algorithms in compressed sensing and distributed compressed sensing. The recovery algorithms include Orthogonal Matching Pursuit algorithm, Bayesian algorithm, L1-norm optimization algorithm and so on.Chapter 3 discusses various propagation characteristics of channels in detail by analyzing the characteristies of wireless communication fading channels. Then the main principle of OFDM technique is introduced. Based on the introduction, common channel estimation algorithms in OFDM systems are researched and analyzed.Chapter 4 first introduces OFDM system model, then gives analysises of current CS channel estimation method. With these studies, new estimation method based on distributed compressed sensing in OFDM and Amplify-and-forward relay systems are proposed separately. Compared with traditional estimation scheme, the proposed methods can greatly reduce the number of pilots, and achieve a good tradeoff between the performance and the complexity by choosing the number of virtual channels. In addition, simulations results are provided in different communication environments, and the performance verifies the correctness of the analysis.Chapter 5 researches on channel estimation schemes by combining CS and Kalman filtering. In this chapter, current Kalman filtering channel estimation methods are first analyzed, and some drawbacks in theses methods are listed. Then two innovation schemes are proposed and complexity comparison of different schemes is given. In the first innovation scheme, by considering the combination of frequency domain Kalman filtering and distributed compressed sensing, the pilot number and the estimation error are both reduced; while in the second innovation scheme, Kalman filtering channel estimation in time-delay domain is studied. CS is introduced to gurantee the correctness of the estimation of multipath delay information. Although both of these two innovation schemes increase the system complexity, the overall complexity can still be accepted in wireless communications, and there are still complexity gains in wideband communication due to the robustness of CS recovery algorithm. Simulation results are provided in the final to verify the correctness of the analysis.Chapter 6 analyzes the theoretical performance of CS channel estimation in low SNR environment. By given new sufficient and necessary condition that wireless channels achieve the energy efficiency of coherent channels, the theory of channel estimation in low SNR regime is perfected. In this chapter, current condition conclusions are first studied, then based on the OFDM transmission model, improved condition with CS channel estimation are derivated and discussed. Compared with current condition, new condition is more suitable for practical application with the help of CS. Besides, when considering other transmission mode, similar conclusions can also be obtained. Chapter 6 concludes this dissertation, and presents some work in the future.