Development of Blind Signal Separation Technologies and their Application to Mobile Communications

盲信号处理是目前信号处理领域中的一个新兴领域，具有重要的理论基础和广阔的应用前景。事实上，BSP已经成为许多领域非常重要的研究和开发课题，特别是在移动通信、声学和语音处理以及生物医学工程方面。BSP技术基本上不使用任何训练数据，并且不假定关于瞬时混合或卷积混合系统的参数的先验知识。在这项研究中，我们结合盲源分离技术在下一代移动通信中的应用，研究了卷积混合的盲信号(或源)分离问题。我们提出了几种盲源分离方法，并通过数字仿真实验考察了所提方法的有效性。粗略地说，所提出的方法可分为以下四类：(1)自适应超指数过程：在线盲源分离…慢时变环境下的更多技术。(2)稳健超指数过程：噪声环境下的离线盲源分离技术。(3)带参考信号的特征向量法：具有非常高的盲源分离成功率的离线盲源分离方法。(4)带参考信号的稳健特征向量过程：噪声环境下的离线盲源分离技术，具有非常高的盲源分离成功率。对于上述(1)和(2)，尽管我们在2000年提出了最初的(多通道)超指数方法，但我们提出了原始方法的自适应版本，可以用于慢时变环境。关于(3)，我们把B.Jelonnek和K.D.Kammeyer在1994年提出的单输入单输出(SISO)系统的带参考信号的特征向量法推广到多输入多输出(MIMO)系统的情形。至于(4)，我们推广了上述特征向量法(3)在噪声环境下的情况，通过以上四种类型的盲源分离程序的开发，为下一代移动通信中的盲源分离奠定理论基础。我们相信，这一理论基础为我们在下一代移动通信中设计先进的信源检索器(或均衡器)提供了原则。较少

Blind Signal Processing (BSP) is now one of emerging areas in signal processing with theoretical foundations and many potential applications. In fact, BSP has become a very important topic of research and developments in many areas, in particular, in mobile communications, acoustics and speech processing, and biomedical engineering. BSP techniques principally do not use any training data and do not assume a priori knowledge about parameters of instantaneous mixing or convolutive mixing systems. In this research, we deal with the blind signal (or source) separation (BSS) problem for convolutive mixtures with taking the application of BSS techniques to next-generation mobile communications in consideration. We proposed several procedures for BSS and investigate the effectiveness of the proposed procedures through digital simulation experiments.Roughly speaking, the proposed procedures are classified into the following four categories:(1) Adaptive super-exponential procedures: On-line BSS … More techniques in slowly time-varying environments.(2) Robust super-exponential procedures: Off-line BSS techniques in noisy environments.(3) Eigenvector procedures with reference signals: Off-line BSS techniques with a very high success rate of BSS.(4) Robust Eigenvector procedures with reference signals: Off-line BSS techniques in noisy environments with a very high success rate of BSS.As for (1) and (2) above, although we proposed the original(multi-channel) super-exponential methods in 2000, we proposed an adaptive version of the original ones, which can be utilized in slowly time-varying environments. As for (3) , in connection to the super-exponential methods, we extended the eigenvector method with reference signals for single-input-single-output (SISO) systems proposed by B. Jelonnek and K. D. Kammeyer in 1994 to the case for multi-input-multi-output (MIMO) systems. As for (4), we extended the above eigenvector method (3) the case in noisy environments.Through the developments of the above four types of BSS procedures, we will establish a theoretical foundation for BSS in next-generation mobile communications. We believe that the theoretical foundation gives us a principle for designing advanced source retrievers (or equalizers) in next-generation mobile communications. Less