Research On Detection Algorithms And Performance Analysis For Spatial Interference Alignment

Improving spectrum efficiency is one of the major goals for designing wireless communication systems. Due to the lack of effective process scheme, signal interference is one of the main limited factors for enhancing spectrum efficiency. Recently, Interference Alignment (IA) has been proposed to solve this problem. The research results show that signal interference is not a fundamental limitation in wireless communications system. In theory, IA can take K/2 multiplexing gain in interference channels, where K is the number of users. This means the capacity of IA system is proportional to the number of users. Therefore, IA attracts considerable research attentions. In IA scheme, multiple interfering signals are aligned into a rank deficient subspace at each receiver so that the desired signal can be derived free of interference.The researches for IA are divided into following four parts, closed-form IA algorithm, iterative IA algorithm, IA with diversity and the IA performance analysed based on random matrix. The main works of this dissertation are summarized as follows:First, the feasible conditions of IA are proposed, and non-uniqueness of IA solutions is proved. In certain cases, the precoding matrix of IA can be obtained directly, namly closed-form IA. However, it is very difficult to get the optimal closed-form IA solution due to multiplicity of its solutions. To conquer this, we elaborate the characters of both the desired signal and the interference subspace imitation of unitary space in I A, and achieve the relationship between their inner product and desired signal power. Furthermore, a Mini-norm of inner product algorithm is proposed. The numerical results show the algorithm can improve both the desired signal power and the sum rate performance.Note that IA precoding matrics can be obtained by closed-form solutions only in some cases. In general, the IA precoding matrics can be obtained only through an iterative algorithm. In iterative IA algorithm, the objective function is proposed and thus the precoding matrices can be derived by minimizing the objective function associated with interference. The classic iterative IA algorithm is a distributed IA. Max-SINR algorithm can be taken as the bound for interference channel to some extent because it does not converge. This dissertation proposes a weighted algorithm that takes both interference and the desired signal into account. The algorithm achieves precoding matrices by minimizing the objective function iteratively. Both the constant weight and the adjustable weight are considered and their convergences are proved. Though the weighted algorithm is inferior to Max-SINR at low SNR, it outperforms the distributed IA. Moreover, the adjustable weight algorithm is superior to the constant weight algorithm.IA focuses on eliminating the influence of interference on the desired signal, which leads to its poor performance at low SNR. In order to solve the problem, this dissertation studies both the methods and performance of diversity IA. To begin with, the conditions of diversity IA are shown. Then taking the receiving diversity for example, the dissertation shows the implement algorithm of single side diversity and its performance. Next, the possibility of achieving diversity at both receiving side and transmitting side at the same time is explored and a preliminary analysis of the results of double side diversity is presented. Numerical results show that the diversity can improve the IA performance at low SNR. However, at high SNR, the performance of multiplexing is better than that of the diversity.To obtain IA performance in complex communication system, it is necessary to get its performance approximation by the theoretical analysis. The theory of random matrix provides a tool for the physical layer of a wireless communication system, as well as for interference channel. First, we achieve the theoretical sum rate performance by analyzing the Asymptotic Eigenvalues Probability Distribution Function (AEPDF) of IA. And then we achiev the same performance based on deterministic equivalent that obtains the solutions by fixed point methods. Both the two methods can predict the IA sum rate performance. The latter has a wider range and is effective in investigating the diversity of IA. Moreover, the method of deterministic equivalent still holds when the limit of channel matrics dose not exist.