关键词: 独立成分分析, 盲源分离, 牛顿迭代法, 功能磁共振成像

Abstract:

As the fixed-point algorithm and the infomax algorithm, two of the most popular algorithms of indepen- dent component analysis （ICA）, spend too much time in processing functional magnetic resonance imaging （fMRI） data, an optimization model of ICA is presented. Based on the model, a fast Newton iteration algorithm is pro- posed, in which an improved Newton iteration method is adopted to achieve a three-order convergence speed. The proposed algorithm and the two above-mentioned algorithms are then used to process real fMRI data. The results show that the proposed algorithm well separates the independent components from fMRI data with less computation and high convergence speed, and that it has obvious advantages in processing fMRI signals with huge numbers of data

Key words: independent component analysis, blind source separation, Newton-Raphson method, functional magnetic resonance imaging