针对FastICA算法存在依赖非线性函数选取的缺陷,为了提高分离结果的可靠性,提出一种基于蚁群算法的改进ICA算法。该算法对非线性函数没有特殊要求,以负熵近似表达式为目标函数,利用蚁群算法代替FastICA算法中的牛顿梯度法,求出最优分离矩阵B,从而对混合信号中的独立分量进行分离。仿真结果验证了改进ICA算法的有效性和优越性。
The FastICA algorithm has the defect in relying on the selection of nonlinear functions. In order to improve the reliability of the separation resuits, an improved independent component analysis algorithm based on ant colony algorithm is introduced. Such algorithm has no special requirements of nonlinear function, takes the approximate expression of negative entropy as the objective function, and can be optimized by taking use of ant colony algorithm instead of Newton gradient method. The best separation matrix is found and then the independent components from the mixed signals are separated. Simulation result proves the improved independent component analysis algorithm is effective and better.