High Dimensional Sparse Regression Model For Channel Estimation In Underwater Acoustic OFDM

The underwater acoustic(UWA)communication performance largely hinges on the channel estimation accuracy at the receiver,at the physical layer,UWA channel pose grand challenges for effective communications.This is because the UWA channel has a long path delay spreads,as well as serious Doppler effects.In single-carrier transmission,long channel delay spread leads to significant inter-symbol-interference(ISI).For multicarrier approaches like orthogonal frequency division multiplexing(OFDM),aforementioned Doppler effects destroy the orthogonality of the sub-carriers and lead to inter-carrier-interference(ICI).The effective way to overcome this difficulty is to use the sparse characteristic of the UWA channel,meaning that most of the channel energy is concentrated in a few path delay and Doppler value.In recent years,the sparse characteristic of the underwater acoustic channel is used,and the wide used UWA channel estimation technology mainly has the following two categories: one is the use of subspace algorithms from the array processing literature,can be applied for channel estimation.The other one is based on compressed sensing(CS)met hod,now is the most popular algorithms for sparse UWA channel estimation,and outperform to the conventional least-square(LS)channel estimator.The existing research shows that the main problem of the current channel estimation is,the above channel estimation methods assume that the known or experimentally obtained Doppler is specific and error free.But in fact,due to medium instability,such as the current-induced platform motion and wind-generated waves as time-varying reflectors,the Doppler in the experiment has a great error to the Doppler value in the real environment.This affects the estimation accuracy of the UWA channel,which leads to the low performance of the UWA communication.At present,the underwater acoustic channel estimation which c an tolerate Doppler error is very limited,and this is also the main reason for this study.We present to solve this Doppler inaccurate issue based on the high-dimensional sparse regression(HDSR)model in statistical learning area.Firstly,the formula transformation is made by using the norm optimization problem of the UWA channel estimation in the mainstream CS method,and then the HDSR optimization problem model is obtained.We proved that the channel estimator obtained by the HDSR model can converge to the true solution of UWA channel;the theory ensures that the HDSR model can be used to estimate the channel accurately in the case of Doppler uncertainty,and improve the performance of UWA communication.Secondly,we give a simple projection gradient de scent(PGD)algorithm based on the HDSR model to generate a series of iterative sequence,and we also proved that the iterative sequence converges to the neighborhood of channel estimator in HDSR model.Finally,based on the simulation results,we observe that the PGD algorithms with HDSR model outperforms the OMP,Spa RSA and BP methods,and the PGD algorithm can more effectively handle channel with significant Doppler spread.