论文部分内容阅读
本文研究了后非线性混合信号的盲分离 .后非线性混合信号是由线性混合的每一路信号分别经过一个非线性畸变产生的 .因此分离这种信号需要在适用于线性混合的线性分离结构前放置一个用于补偿非线性畸变的非线性校正部分 .本文用一种最大似然方法推导了一般后非线性分离结构的学习公式 .在前人一些工作的基础上 ,提出了一种用于亚、超高斯信号后非线性混合的盲分离算法 .该算法用多层感知器对分离结构的非线性校正部分进行建模 ,迭代过程中根据一稳定性条件在分别适用于亚、超高斯信号的概率模型间进行切换并以块自适应方式工作 .通过对模拟信号及实际信号 (图像和语音 )的实验证明了该算法的有效性 .
In this paper, we study the blind separation of the post-mixed nonlinear signals, which are generated by a nonlinear distortion of each linearly mixed signal, so it is necessary to separate the signals before the linear separation for linear mixing A non-linear correction part for compensating nonlinear distortion is placed in this paper.A maximum likelihood method is used to derive the learning formula of general post-nonlinear separation structure.On the basis of some previous work, , And a non-linear mixed blind separation algorithm based on Gaussian signal.The proposed algorithm uses a multilayer perceptron to model the non-linear correction of the separation structure. The iterative process is based on a stability condition for the sub-Gaussian signal The probabilistic models are switched and work adaptively in blocks.Experiments on the analog signals and real signals (image and speech) prove the effectiveness of the algorithm.