摘要在过去的十多年间，多输入多输出(MIMO ，Multiple-input Multiple-output)技术被广泛的关注和研究，它能够在不增加带宽的情况下大幅度地提高系统的信道容量和频谱利用率，因此该技术也被视为未来无线通信的关键技术之一。

本文首先对最大似然检测（ML，Maximum Likelihood），迫零检测（ZF，Zero Forcing），最小均方误差检测（MMSE，Minimum Mean Square Error），连续干扰消除检测（SIC，Successive Interference Cancellation），排序的连续干扰消除检测（OSIC，Order Successive Interference Cancellation）和正交三角（QR，Orthogonal Triangle）检测等传统算法进行分析，通过仿真可知ML检测性能最优，但复杂度过高，ZF/MMSE/QR检测性能最差，但复杂度也是最低，干扰消除算法的性能和复杂度介于两者之间。66663

随后对接近于最大似然检测的球形检测（SD，Sphere  Decoding)算法进行初步研究，基于深度优先的球形检测算法在高信噪比（SNR ，Signal to Noise Ratio)时复杂度很低，但在低信噪比时却依然接近于最大似然的指数数量级的复杂度。而基于宽度优先的K-best检测算法可以通过改变保留的节点数来获得复杂度和误码率（BER ，Bit Error Rate）的良好折衷。

最后对半定松弛 (SDR，Semi-definite  relaxation) 检测算法进行初步研究，由于最大似然检测算法复杂度过高，通过放宽最大似然检测的约束条件，将非凸的优化问题近似为凸规划的问题来解决，这样将大大降低计算复杂度。

毕业论文关键词：多输入多输出，传统检测算法，最大似然算法，球形译码，K-best，半定松弛

毕业设计说明书（论文）外文摘要

Title High-performance  Detectors  for  MIMO  Communication

Systems

Abstract

In the past decade, multiple input multiple output (MIMO) technology has attracted wide attention and heavy research, it can greatly increase channel capacity and spectrum efficiency  without increasing bandwidth of system, so  it is also  regarded as one of the key technologies in the  future  wireless communications.

Firstly, several traditional MIMO detectors are analyzed and studied as follows:  maximum likelihood (ML), zero-forcing (ZF), minimum mean square error (MMSE), successive interference cancellation (SIC), ordering SIC(OSIC) ，orthogonal triangle (QR) and other typical detection algorithms. Their performances are compared by simulation. From simulations, we can know that the ML detector has the optimal performance, but its complexity is so high that it cannot be realized for the case of a large number of transmit antennas or a large signal constellation size. All ZF/MMSE/QR detectors have the worst performance compared to ML and SIC, however their complexities are the lowest. SIC makes a good compromise between complexity and performance.

Then, sphere decoding (SD) algorithm is investigated, whose performance is close to ML. The complexity of the  SD  algorithm based on depth-first is very low in high signal-to-noise ratio (SNR) region, but its bit error rate （BER） performance is still very close to the maximum likelihood with exponential complexity in low SNR region. And the K-best SD detector based on breadth-first can change the number of nodes to strike a balance between complexity and BER.

Finally, we would do a preliminary study of the semi-definite relaxation detection because the SD has a very high complexity at low SNR region. By relaxing the constraint of maximum likelihood detector, a non-convex optimization problem can be converted to a convex programming problem, this processing will greatly reduce the computational complexity when preserving or keeping a good BER performance.

Key words: multiple input multiple output, the traditional detection algorithm, the maximum likelihood algorithm, sphere decoding, K-best, semi-definite relaxation