单通道混合信号多尺度分析与盲源分离

项目名称： 单通道混合信号多尺度分析与盲源分离

项目编号： No.61301250

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 郭一娜

作者单位： 太原科技大学

项目金额： 26万元

中文摘要： 单通道盲源分离具有独特的优势，已成为生物医学、海洋环境、声学监测和军事侦察领域中的一个热点课题。传统的盲源分离法要求观测信号的数目不少于源信号数目，而在实际应用中受造价和安装条件等因素限制，常使观测信号的数目远少于源信号数目，从而传统盲源分离法很难恢复出源信号。因此，如何凭借单通道混合信号恢复出多通道源信号是数学领域中的一个具有挑战性的课题。 本项目拟研究在源信号理论模型和信道参数无法精确获知的情况下，仅根据单路混合信号便可恢复出多路源信号。首先分析各种单通道混合信号特征并对其分类建模，然后针对已知的单通道混合信号模型及参数进行自适应分离并恢复多路源信号。本项目依据各种信号的统计特性、时频域特性和非线性特性，提出多尺度相结合的单通道盲源分离法，并发展传统单通道盲源分离理论。该方法可提高源信号复原性能，具有算法复杂度低、收敛速度快及鲁棒特性强的特点，可解决一系列单通道盲源分离法的瓶颈问题。

中文关键词： 盲源分离；单入多出；多尺度分析；时频域分析；非线性分析

英文摘要： Single channel blind source separation has become a hot topic in the fields of biomedical sciences, marine environment, acoustic monitoring and military reconnaissance with its unique advantages. Traditional blind source separation requires the number of observation signals not less than the number of source signals. However, in reality, the number of observation signals is always less than the number of source signals for the limitations of cost and installation conditions. Thus it is of great difficulty to recover source signals by use of traditional blind source separation methods.Therefore how to recover multi-channel source signals from single channel mixed signal is a challenging subject. This project studies the recovery of multiple channel signals only by single channel mixed signal under the situations that source signal theory model and channel parameters cannot learn accurately. At first, all types single channel mixed signals are analyzed, classified according to their properties. Then multi-channel source signals are separated and recovered adaptively based on the known single channel mixed signal model and parameters. The ideology presents a multi-scale combinated single channel blind source method according to statistical properties, the time and frequency domain characteristics and nonlinear cha

英文关键词： Blind source separation；single input multiple output；multiscale analysis；time-frequency analysis；nonlinear analysis