Abstract: In the broadband wireless communication system, the methods of adaptive sparse channel estimation can improve the accuracy of the channel estimation since it can make full use of the inherent sparseness of the channel. However, most of traditional sparse channel estimation methods are based on the hypothesis of Gaussian noise models, the channel estimation performance will be severely deteriorated when environmental noise contains impulsive noise. To deal with this problem, a robust variable step-size sign algorithm based on mixed norm of gradient vector was proposed. Simulations demonstrate that the proposed algorithm achieves the expected performance in Gaussian noise and impulsive noise environment.

Key words: sparse channel estimation, impulsive noise, variable step-size, sign algorithm