Abstract
Interference alignment(IA) was proposed recently for suppressing interference to achieving the optimal degree of freedom of an interference channel. For reducing the complexity of IA algorithms, we present two modi ed IA algorithms in this thesis. The first IA algorithm is called the mixed IA(MIA) algorithm, which used diff erent criteria to minimize the leakage interference and mean squared error, for designing the precoding matrices and combining matrices, respectively. The MIA algorithm provides lower complexity than the maximum signal-to-interference-plus-noiseratio(MSINR) algorithm or the iterative interference alignment(IIA) algorithm, while demonstrating comparable performance. This algorithm results in orthonormal precoding matrices, which possess the advantage of efficient quantization. On the basis of the MIA algorithm, we also propose a weighted mixed IA(WMIA) algorithm, which accounts for the weak user's performance. The WMIA algorithm manages interference by di fferent weights to distribute power fairly and it also possesses the advantages of low complexity and orthonormal precoders, similar to the MIA algorithm. However, it demonstrates trade-off performance compared with that of the MIA algorithm in diff erent situations. The other issue investigated in the thesis is initialization. As IA algorithms do not guarantee to achieve the global optimum; hence the initialization is crucial for the performance. We propose an initialization method with low complexity that involves maximizing the desired signal power through the gradient descent method. In this study, we consider the multi-user multiple-input multiple-output orthogonal frequency division multiplexing(MIMO-OFDM) interference system. The channel model employed is the 60GHz indoor channel model established by the IEEE802.15.3c task group, which includes line-of-sight(LOS) and non-line-of-sight(NLOS) situations.