Matrix analysis
Natural gradient works efficiently in learning
Neural Computation
Equivariant nonstationary source separation
Neural Networks
Hi-index | 0.08 |
Multichannel blind equalization is an important task for numerous applications such as speech separation, dereverberation, communication, signal processing and control, etc. In this paper, only second-order statistics is used to construct a cost function and thus a new online algorithm is derived with natural gradient search method for multichannel blind equalization. The proposed algorithm can deal with observations with lower-level noise. Simulations and comparisons indicate the efficiency and ability of the algorithm to perform blind equalization with simple calculation, fast convergence speed and high accuracy, and also can carry out speech separation and dereverberation.