Model-Data Fused Acoustic Channel Equalization Under Fluctuating Shallow Water Environment

Underwater acoustic communication achieves wireless data transfer in the ocean via propagation of sound. It faces numerous technical challenges different from radio communications in air, due to special channel features such as limited bandwidth and relatively slow propagation speed. Particularly in shallow water environments, fluctuations along the boundaries and in the media cause severe multipath time-delay spread and Doppler frequency-shift spread, both of which can be rapidly time-varying, resulting in poor performance or even failure of conventional high-rate communication techniques.Facing the above-mentioned doubly-spread and rapidly time-varying channels, this thesis adopts an approach that takes into account environmental variations, sound propagation channel, and communication signal processing all together. Especially, time-reversal processing exploiting sound propagation physics is incorporated into coherent single-carrier and multi-carrier underwater acoustic communications, from which some model-date fused channel estimation and equalization techniques are developed. As such, we aim to not only obtaining deep understanding of the physical model of the shallow water communication signal propagation, but also improving signal processing methods and system design for reliable high-rate acoustic data transfer in realistic fluctuating environments.The channel impulse response (CIR) of shallow water acoustic communication differs from place to place partially because sound speed profiles follow different patterns due to the temperature and salinity variations. Currently there is no universal CIR model that works in all the shallow water conditions. That requires a combination of model and data, which takes advantage of both the models' generality and data's specialty. Time reversal is a signal processing method developed based on the reciprocity of sound propagation in a time-invariant environment. Specifically in the context of array receiving, when the environment is stationary, if the received signals at the array from initial transmission are time reversed and re-transmitted, one can obtain the original signal at the original transmitting position via spatial focusing and time compression. This thesis studies relevant aspects of the time-reversal processing both theoretically and experimentally, and apply it to coherent underwater acoustic communications. In a fluctuating shallow water environment, however, time-varying CIR is the main factor causing communication performance degradation. This motivates the thesis research to study the influence of the environmental fluctuation on communication channel and then track the CIR variations. We first develop a state space time-varying model for a channel with moving scattering clusters:we then apply the standard Kalman filter and extended Kalman filter to estimate and track the channel CIR. Since the CIR has a large time delay spread and changes with time, the required computational effort is significant. Thus we propose to track the so-called time-reversed channel (TRC), which is the channel correlation function after applying time-reversal processing. Theoretical analysis and data observation show that TRC has a smaller time-delay spread and more stable structure compared to CIR. Hence tracking the time-reversed channel requires less computation as well as lower update frequency.Understanding the characteristics of the time-reversed channel provides a basis for equalizer design in passive time reversal communications. Due to time-varying, time-reversed channel does not show as an ideal impulse response; instead the response still has some spread, i.e., there still exists some residual inter-symbol interferences (ISI) in communications. Hence passive time reversal processing can be followed by a simple decision feedback equalizer to remove the residual ISI. As the channel varies with time more rapidly, the time-reversed channel will spread even more; it is thus necessary to update the CIR used to implement the passive time reversal algorithm. In this thesis, errors in Kalman filtering are used to track the variation of the time-reversed channel, i.e., if the predicted error goes beyond a predetermined threshold, the CIR used for passive time reversal is then updated. Experimental data processing and analysis verify that the proposed approach is feasible.This thesis also investigates the use of single-carrier frequency domain equalizer (SC-FDE) in underwater acoustic communications. Having disclosed that the mathematical expression of the SC-FDE is quite similar to that of passive time reversal combined with temporal decision feedback equalizer, we compare those two methods from both theoretical and experimental data perspectives. It is shown that when the environment is stationary, they perform equally well. In a dynamic environment. SC-FDE needs to update the CIR more frequently. Besides. implementation of the SC-FDE is based on fast Fourier transform which requires a cyclic extension for each transform data block to warrant periodic extension of data sequences; further length of the cyclic prefix has to be longer than the CIR spread. This signal sequence formation reduces the average communication data rate.Given the favorable features of the time-reversed channel, we further exploit processing the passive time reversal incorporated into the Orthogonal Frequency Division Multiplexing (OFDM) system. Considering the smaller delay spread of the time-reversed channel, we can design shorter OFDM symbols so that the sub-carrier can have a larger bandwidth; the combined OFDM system can thus better handle inter-carrier interferences. Simulations and experimental data analysis demonstrate the effectiveness of the scheme.