Generalized Gauss density model was employed for the blind separation of the mixtures of sub- and super-Gauss sources, and the local stability analysis was performed for independent component analysis algorithm in the framework of the natural gradient in Stiefel manifold. We presented some new results about its local stability analysis and verified their correctness and applicability.%基于广义Gauss概率密度模型对超Gauss和亚Gauss源混合信号进行盲分离,在Stiefel流形的自然梯度法框架下,分析了独立成分分析算法的局部稳定性,并采用几种典型非线性函数验证了所得结论的正确性和适用性.
展开▼