变步长盲源分离算法的研究-信号与信息处理专业论文.docxVIP

I 摘要 盲源分离技术是近年来信号处理、神经网络及智能计算等领域中新涌现出来 的热点问题之一，其在地震勘探、无线通信、医学成像、语音增强、生物医学、 计量经济学等方面都有着重要的应用价值。本文详细阐述了盲源分离的基本理 论，研究了自适应及批处理的盲源分离问题，并讨论了步长与自适应算法分离性 能间的关系，提出了一种新的变步长盲分离算法。主要内容概括如下： 本文首先介绍了盲源分离的基本知识及相关理论基础，阐述了盲源分离的两 大类型的处理方法，着重分析了自适应盲源分离算法的理论依据、算法结构以及 其步长参数如何进行优选的有效策略；然后，文中提出了一种新型的变步长 EASI 盲源分离算法，该算法通过参考自适应滤波技术中的步长调节方法，利用设定分 离矩阵权范数来描述 EASI 算法所处的分离状态，并根据该参数呈现指数趋势下 降的变化规律来进一步获得步长参数的自适应更新规律。仿真实验表明该算法有 效地加快了分离算法的收敛速度，减小了稳态误差，且对于时变混合环境具有一 定的实时跟踪能力；最后对全文进行了总结并就盲源分离问题的未来发展趋势进 行了展望。 关键词：盲源分离，变步长，自适应，EASI II Abstract In recent years, Blind source separation technology in signal processing, neural networks and computational intelligence in areas such as emerging out of one of the hot issues, and its Seismic exploration, wireless communications, medical imaging, speech enhancement, biomedical, and other aspects of econometrics has important application value. This paper describes the basic theory of blind source separation, adaptive and batch processing of blind source separation problem and discuss the adaptive algorithm step size and the relationship between separation performance, a new variable step size blind separation Algorithm. Mainly be summarized as follows: This paper introduces the blind source separation based on the basic knowledge and related theories to explain the two types of blind source separation approach, focusing on analysis of adaptive blind source separation algorithm theoretical basis, the algorithm structure and how its parameters step Selection of an effective strategy; then, this paper proposes a novel variable step size EASI algorithm for blind source separation, the algorithm adaptive filtering techniques by reference to the step adjustment method, the use of the right of separation matrix norm set to describe the EASI algorithm in which the separation of state and exponentially according to the parameter variation of the decline of steps to further access to the adaptive parameter update laws. Simulation results show that the algorithm