Neural blind deconvolution of MIMO noisy channels

In this paper blind deconvolution of multiple-input-multiple-output channels under a noisy environment, is considered. The noisy signals are modeled by a finite-impulse response filter and zero-mean Gaussian signal. After cancellation of noise by using the adaptive learning algorithm, a fully connected Herault-Jutten network with delays, is used to perform blind deconvolution. The proposed algorithm can be implemented for online operation and is capable of delivering a very consistent performance. Obtained results corroborate the effectiveness of the proposed algorithms.