Research On Underwater Acoustic Channel Tracking Method Based On Paths Correlation

The acquisition of underwater acoustic channel state information is the basis of underwater acoustic communications.Due to the dynamic time-varying characteristics of the underwater acoustic channels,the pure channel estimation technique cannot meet the requirements of underwater acoustic communications.Underwater acoustic channel tracking is a technique for dynamically tracking channel state information.The research of underwater acoustic channel tracking is of great significance to the improvement of underwater acoustic communication performance.Existing studies have shown that paths in many underwater acoustic channels have certain cross-correlation properties.Making full use of this feature can contribute to the improvement of underwater acoustic channel tracking performance.According to this characteristic,the high-dimensional channel can be decomposed into channel principal components and channel subspaces with low dimensions for tracking,which improves the tracking performance and reduces the computational complexity.For the channel subspace,we propose a channel reconstruction tracking method based on orthogonal subspace tracking method.The subspace tracking algorithm used in the existing channel reconstruction tracking algorithm obtains that the channel subspace does not have orthogonality,which will result in inaccurate channel principal component extraction,which in turn leads to a decrease in channel tracking performance.We improve the subspace tracking method to maintain the channel subspace orthogonality during the tracking process,and realize the tracking of the channel principal component,and then reconstruct the channel state to realize the tracking channel.The ocean experimental data effectively proves that the channel reconstruction tracking method based on the orthogonal subspace tracking method proposed above improves the channel tracking performance in both calm sea and rough sea conditions.For the channel principal component,we propose a training-based adaptive channel tracking method and a forward-backward Kalman channel tracking method.Both methods firstly model the channel principal component as an autoregressive(AR)process.Then,based on this model,the Kalman filter algorithm is used to track the channel principal component.However,this model mismatch is inevitable under water.Moreover,there are many parameters to be determined involved in the use of the model,and the channel state changes continuously with the fluctuation of the water.These will cause the model to be mismatched.Therefore,we propose a training-based adaptive channel tracking algorithm that adaptively corrects prediction errors by using an adaptive Kalman filter.The parameters of the tracker are obtained through the training data.Furthermore,we propose a forward-backward Kalman channel tracking algorithm to improve the tracking of channel principal components by using the measured values of forward and backward moments,and to compensate for the tracking performance loss caused by model mismatch from another angle.Through ocean data verification,the two channel tracking methods proposed above have lower signal prediction errors than traditional methods,and the tracking performance is greatly improved.Experiments show that the above method for channel subspace is more practical for the relatively calm situation of underwater transmission environment.The second type of channel principal component method is more practical for the relatively rough situation of underwater acoustic transmission environment.