An optimization model for independent component analysis (ICA) is built. A new gradient algorithm based on the model is presented, which is called Orth-ExtBS algorithm. This algorithm combines the advantages of the FastICA and ExtBS algorithms, which is easy to use in simple form and is able to separate blindly mixed signals with sub-Gaussian and super-Gaussian source distributions. The accuracy of the Orth-ExtBS algorithm is high and its convergence speed is fast. Applying the Orth-ExtBS algorithm and two other algorithms (FastICA and ExtBS) to fMRI data, the results show that the new algorithm is superior to the other two on accuracy of estimating the temporal dynamics of activations by comparison.
|Number of pages||5|
|Journal||Dalian Ligong Daxue Xuebao/Journal of Dalian University of Technology|
|State||Published - Jul 2005|
- Blind source separation
- Functional magnetic resonance imaging
- Gradient algorithm
- Independent component analysis