论文部分内容阅读
针对具有二阶非平稳特性的源信号盲分离问题,提出一种基于自组织神经网络的在线盲源分离新算法.利用自组织神经网络构建一种多层盲分离网络模型,以网络输出层信号的相关性为代价函数,采用自然梯度原理对网络参数进行学习,最小化该代价函数从而实现信号分离.将多层自组织神经网络和自然梯度原理相结合,提高了分离算法的灵活性和性能.最后将该算法与其他算法进行了仿真对比,仿真结果表明该算法具有较好的收敛精度及稳定性.
Aiming at the problem of blind separation of source signals with second-order non-stationary characteristics, a novel online blind source separation algorithm based on self-organizing neural networks is proposed. A multi-layer blind separation network model is constructed by using self-organizing neural networks, , The natural gradient principle is used to study the network parameters to minimize the cost function to achieve signal separation.The combination of multi-layer self-organizing neural network and natural gradient principle improves the flexibility and performance of the separation algorithm Finally, the algorithm is compared with other algorithms and the simulation results show that the algorithm has better convergence accuracy and stability.