单通道同频混合信号的盲可分离性、性能界及算法研究

Single-channel blind separation of co-frequency mixed signals is one of the research hotspots and difficulties in the signal processing area. Now there are two mainstream blind separation algorithms: one is joint parameter estimation and sequential detection, another is based on bit-interleaved coded modulation with iterative decoding(BICM-ID). The former algorithms' computation complexity are growing exponentially with the increase of modulation order and channel order, so it is hard to apply to high order modulation signals. The latter algorithms' performance can be declined sharply for high code rate signals. This project would start with the blind separability of single-channel co-frequency mixed signals, and search for the separability conditions(or parameter characteristic). When the mixed signals are separable, the Cramer-Rao lower bound of each parameter would be researched by multi-parameter inequation, then parameter estimetion algorithms wound be studied further, which can close to the performance bound. Then the separation performance lower bound will be taken into cosideration from the point of information theory, in which Fano inequation is utilized. Finally, blind separation algorithm with modest computation complexity will be researched, which not only can be applied to high order modulated signals, but also can be close to lower bound of performance. the project aims at showing the separability law of single-channel co-frequency mixed signals( which are overlapped in time domain, frequency domain and space domain) and searching for separation performance bound as well as separation algorithms, which can establish a theoretical basis for the blind separation of single-channel mixed signals.

单通道同频混合信号的盲分离问题是近年来信号处理领域研究的热点和难点。目前主流的盲分离算法有两类：一是联合参数估计与序列检测的分离算法；二是利用比特交织编码调制迭代译码分离算法。第一类算法的复杂度随调制阶数和信道阶数的增加呈指数增长，难以应用于高阶调制信号；第二类算法在应用于码率较高的信号时性能急剧下降。本项目拟从单通道同频混合信号盲可分性入手，研究混合信号可分的条件（或参数特征）；在满足可分离性的条件下，利用多参量Cramer-Rao不等式，研究混合信号各个参数的Cramer-Rao下界，进而研究接近性能界的参数估计算法；之后从信息论的角度，利用Fano不等式给出盲分离的性能下界，研究接近性能界且复杂度适中、适用于高阶调制信号的盲分离算法。项目旨在阐明单通道同频混合信号（即时域、频域和空域都混叠的信号）可分性的规律，并找到其分离性能界和分离算法，为单通道混合信号的盲分离研究奠定理论基础。

随着诸如成对载波多址（Paired Carrier Multiple Access，PCMA）等先进通信体制在卫星通信中的广泛使用，同一频带内接收到多个信号分量的可能性大大增加。这些信号分量在时域和频域均发生混叠，有的甚至连扩频码字都完全相同，因此在接收端必须面临混合信号的处理问题。特别地，对非合作接收方而言，很多情况下只能利用单通道处理技术对下行链路中接收的混合信号进行解调。相比多通道接收机，单通道接收机系统简单，成本较低，灵活性好，具有更好的应用前景。因此，非合作卫星通信信号接收中同频混合通信信号的单通道处理技术成为近年来信号分析领域的研究热点。.本文基于联合信道参数估计和序列检测的方法对单通道混合信号进行处理，主要的工作和创新点如下：.(1) 推导了单通道接收同频混合通信信号信道参数估计的MCRB。首先针对两路信号的频偏、初相和时延等六个参数，分六个参数联合估计、参数对联合估计以及单一参数估计三种情况，分别推导出各个参数估计的MCRB，并对不同情况下的MCRB进行对比分析。对两路信号的幅度参数而言，分别推导了联合幅度对参数估计和单一幅度参数估计的MCRB。.(2) 针对具有固定帧长和固定帧同步序列的同频混合信号，研究了可适用于8PSK调制混合信号的频偏、幅度、初相和时延估计算法。从最大似然的角度构造出关于频偏参数的目标函数，进而采用分级搜索的方法获得其最优值。在完成频偏估计的条件下，通过提取按帧累加后帧同步位置待估计参数的信息，实现对幅度、初相和时延三个参数的估计。.(3) 提出了基于联合双向延迟判决反馈序列估计（Joint Double Direction Delayed-Decision Feedback Sequence Estimation，JDD-DDFSE）的混合信号分离算法。利用联合维特比算法（Joint Viterbi Algorithm，JVA）对两路信号分量进行两次联合序列估计，第一次估计利用阶数较短的信道模型对混合信号做初始分离，并将结果进行反馈，确保在第二次JVA中，能对执行JVA过程所用信道抽头外的因果和非因果抽头造成的码间干扰进行补偿。数值仿真表明，对两路QPSK调制下的混合信号而言，当Eb/N0 ≥ 10 dB且反馈的误比特率（Bit Error Rate，BER）小于0.14时，JDD-DDFSE算法是有效的。