基于盲源分离的无线环境感知和分布式信息处理研究

The future wireless networks are facing a dilemma due to the limitation in the frequency resources and the increasing demand of wireless communication services. Aiming at realizing the increasing communication requirements, the intelligent wireless network will be investigated to arrive at through combining with artificial intelligent technology. Because of the inherent technology superiority of blind source separation (BSS), it can separate source signals just from the observed signals without the information of source signal and channel state. Comparing with the traditional signal processing methods, BSS has better power efficiency and spectral efficiency. Therefore, the BSS will be applied to solve the existing problem of spectrum access and interference cancellation in wireless network for purpose of achieving full duplex cognitive radio and low-cost distributed information processing. Firstly, cooperative spectrum sensing algorithm based on BSS will be investigated to overcome hidden primary user problem due to the influence of channel fading and interferences. Secondly, full-duplex (FD) spectrum sensing method will be explored to deal with static sensing and remnant self-interference problems. Finally, distributed BSS information processing will be developed to fulfil the lost-cost and real-time processing requirements. Taking into account the above-mentioned research contents, we will put forward effective models, theories and methods to complete theoretical analysis and simulation verification. This project will contribute significantly to providing theoretical significance and application value, and it can play an indispensable role in favoring technological support for realizing intelligent information processing for future wireless network.

针对未来无线网络面临频谱资源稀缺和日益繁重的通信业务问题，研究结合人工智能理论实现智能无线网络，满足不断增长的通信需求。由于盲源分离具有无需源信号和信道状态信息，仅从观测混合信号分离出源信号等固有的技术优势，与传统的接收信号处理方法相比，盲源分离具有更高的功率效率和频谱效率。本项目针对无线网络中频谱接入和干扰消除问题，将盲源分离用于实现全双工认知和低功耗分布式信息处理。首先，研究基于盲源分离的协作频谱感知算法，解决由信道衰落和干扰引起的主用户隐藏问题。其次，研究基于盲源分离的全双工频谱感知方法，解决静态感知和残余自干扰问题。最后，研究分布式盲源分离信息处理，实现低功耗和实时处理需求。针对上述研究内容，提出有效的模型、理论与方法，完成理论分析和仿真验证。本项目具有重要的理论意义与应用价值，为未来无线网络实现智能信息化处理提供技术支持。

随着无线数据量的急剧增加，通信系统将面临越来越拥挤的频带以及频谱干扰问题，研究结合无监督学习的盲源分离实现智能无线网络处理，具有重要的意义。盲源分离能够在未知源信号先验信息、干扰和传输信道参数情况下有效分离/提取源信号，且不占用通信系统资源，为应对通信环境的复杂性和对抗性带来了新的思路和方法，能够实现无线通信系统增强的接收性能。本项目针对无线网络中频谱接入和干扰消除问题，将盲源分离用于实现全双工认知和低功耗信息处理。项目的主要研究内容包括：研究了增强性能的盲源分离方法用于频谱检测，解决了由信道衰落和噪声影响的性能下降问题，融入通信性能指标到盲源分离的代价函数中，增强频谱干扰消除的有效性；研究了基于盲源分离的全双工频谱感知方法，解决了静态感知和残余自干扰影响问题，增加了系统的频谱感知性能和频谱效率；研究了基于机器学习的协助频谱感知方法，提出了一种基于概率向量的混合AdaBoost方法解决主用户隐藏的问题，实现了性能的有效提升；研究了盲源分离在物联网中解决RFID同频干扰防碰撞的问题，提出了一种欠定模型接收条件下的分离算法，可以使系统的吞吐量提升2倍。 针对上述研究内容，提出了有效的模型、理论与方法，完成了理论分析和仿真验证。本项目具有重要的理论意义与应用价值，为未来无线网络实现智能信息化处理提供技术支持。