Digital signal processing (3rd ed.): principles, algorithms, and applications
Digital signal processing (3rd ed.): principles, algorithms, and applications
Natural gradient works efficiently in learning
Neural Computation
IWANN '01 Proceedings of the 6th International Work-Conference on Artificial and Natural Neural Networks: Bio-inspired Applications of Connectionism-Part II
Hi-index | 0.00 |
In this paper, a novel Frequency-Domain Independent Component Analysis (ICA-F) approach is proposed to blindly separate and deconvolve the convolutive combinations of digitally modulated signals in wireless communications. This approach relies on the simple observation that if signals are independent in one domain, their corresponding components in a linearly transformed domain are also independent. The proposed ICA-F lends itself to computationally efficient Fast Fourier Transform (FFT) implementation, which converts the convolutive combination in the time domain into multiple instantaneous combinations in the frequency domain. Then, the natural-gradient Independent Component Analysis (ICA) algorithm is employed in each frequency bin to the separate frequency components of source signals. The permutation and gain ambiguities associated with the ICA algorithm are successfully solved. The ICA-F has lower computational complexity and faster convergence than the existing time-domain approach. Simulation results confirm the effectiveness of the proposed ICA-F.