Abstract:  Shallow sea is a difficult channel for acoustic communications. Noise in shallow sea acoustic communications may be described by the S S distribution, and the channel modeled as a sparse double selective fading channel. The H1 adaptive filtering is specially designed for none-Gaussian noise, and therefore can be used for channel estimation in an S S noise environment. This paper proposes an algorithm to solve the channel estimation problem based on the H1 adaptive filtering. In addition, to solve the performance degradation problem due to serious individual pulse noise, a signal preprocessing method is proposed. The results show that the performance is improved because of the preprocessing. It is shown that the performance of the H1
adaptive filtering is better than sIPNLMS both in a sparse multipath channel and in a sparse double selective channel.

Key words:  H-infinity adaptive filtering, SAS distribution, doubly selective fading, sparse channel estimation, shallow sea acoustic communication