Research On Channel Estimation And Equalization Of Underwater Acoustic OFDM System

With the progress of science and technology, the demands for marine rescources and exploration will be more and more stronger, which need the underwater information transmission system. The available bandwidth of underwater acoustic communication is small, which is from a few k Hz to tens of k Hz. And at the same time, the future exploration of marine information need high rate underwater acoustic communication system. And the single carrier underwater acoustic communication system can only supply the communication rate at the order of kbps. OFDM technology can greatly improve the spectrum effiency and the communication rate can be increased by one order of magnitude, which is benificial to the high speed underwater acoustic communication target. The underwater acoustic channel is divided into deep sea underwater acoustic channel(depth greater than 100 meters) and the shallow water acoustic channel(depth less than 100meters). At present, the demand for the exploration of shallow water is much greater than that of the deep sea. Hence the thesis pays attention to the shallow water acoustic communication. Shallow water acoustic channel is a multipath delay and Doppler spread channel. The traditional channel estimation technology is established under the dense channel assumption, and the severe spread property requires a lot of pilots. However, a variety of underwater experiments have verified the underwater acoustic channel is a sparse channel. With the prior information, the compressed channel estimation method requires less pilots.However, the underwater compressed channel estimation has two problems.The first one is that the compressed channel estimation model is based on the assumption that the channel is sparsely distributed on the discretized delay and Doppler grids. But the channel actually has sparse distribution on the continuous delay and Doppler domain. The assumption deviation degrades the estimation performance, whcih is called basis mismatch problem. The second one is that the columns of measurement matrix is huge due to the serious spread channel property.And the complexity of compressed recovery algorithm is decided by the product of the measurement matrix transpose and vector. Hence the complexity of the compressed channel estimation is relatively large, which needs the low complexity recovery algorithm. Based on the channel estimation results, the equalization with good performance is needed to ensure the underwater acoustic communication system performance. The recently proposed posterior probability equalization algorithm achieves the performance better than that of the other algorithms.However, the algorithm complexity is too high to the project implementation. A low complexity version of the posterior probability equalization is required while the performance just with a slight decrease.Based on the state of current researches, the thesis majors the above three problems.Firstly, the thesis establishes the basis mismatch model, and based on the model to analysize the effects of basis mismatch on the estimation performance. It indicates that the estimation performance is severely decreased due to the basis mismatch. Recently, the most widely used compensation algorithm is based on the first-order linear approximation to the basis mismatch in order to obtain the engineering realization with convex optimation problems. However the loss of higher order terms brings great performance loss, especially when the basis mismatch is serious. Based on the drawback, the thesis proposes the iterative convex optimization compensation algorithm by alternating iterative recovering the mismatch vector and the sparse channel vector. And then the convergence of the algorithm is analyzed. The thesis also gives the simulations and discussions, and finally comes to the conclusion that the proposed algorithm can effectively improve the estimation performance compared with the other algorithms.Secondly, based on the analysis of the high complexity of the recovery algorithm, the thesis comes to the conclusion that the high complexity is due to the product of the measurement matrix transpose and vector. Based on the conclusion,the thesis proposes block FFT algorithm by making full use of the inherent structure of measurement matrix and the pilot distribution structure. By using the structures,the algorithm divides the computations into severial groups, and in each group with FFT to save the repeate compuations. Besides, the thesis has discussed various special cases about the block FFT algorithm. Then it discusses the complexity and performance of the algorithm, and finally comes to the conclusion that the proposed block FFT algorithm can greatly reduce the complexity and at the same time the performance is not decreased.Thirdly, the thesis has analyzed the high complexity of the posterior probability equalization algorithm, and comes to the conclusion that the high complexity is due to the caculation of the posterior probabilities of all the remaining subcarriers in every iteration. Based on the reason, the thesis proposes the partial posterior probability update equalization algorithm to reduce the complexity by using the banded property of the frequency domain channel matrix, namely, the posterior probability of the subcarriers which has a large distance from the estimated subcarrier of the previous iteration need not to be reestimated. Then the thesisanalyzes and discusses the performance and the complexity, and finally comes to the conclusion that the proposed algorithm largely reduces the complexity and at the same time the performance is only slightly decreased.