Research On Channel Estimation For Underwater Acoustic OFDM Communication Sparse Channel Basedon Compressive Sensing

With the rapid development of wireless communications, people in today's world can't not live without this technology. The environment of the radio channel has become complex. Channel estimation will directly affect the the equality of signal processing and the demodulation of the received signal. Conventional channel estimation method can not take full advantage of the inherent channel sparsity dispersion caused by the dispersion of space, the accuracy and efficiency of channel estimation subject to certain restrictions. Compressed sensing is a new data acquisition processing method; it can accurately recover the original sparse data from a handful of observations in the sample.Sparse underwater acoustic communication channel is one of the common wireless channel. Underwater acoustic communication channel is a time varying multipath channel. the main physical features are Delay spread, Doppler spread, the sound transmission loss, refraction, dispersion, time and space varied noise are the main physical features in the underwater acoustic communication channel.The main parameters of sparse underwater acoustic communication channel based on rrthogonal frequency division multiplexing underwater acoustic communication channel model are analyzed in this thesis. What is more, the performance of the traditional least squares estimation and basis pursuit algorithm on this channel model has been studied. Moreover, the multiple measured vector model based on the channel change is presented, and is compared to the transmit signal cycle. It is considered the channel does not change in a few symbol period, more measurement vector model is proposed to improve the channel estimation accuracy.The orthogonal matching pursuit algorithm on channel estimation for sparse underwater acoustic communication is also studied in this thesis, and an improved algorithm is proposed, which the atoms are selected from the most relevant one. This algorithm can reduce the number of iterations and the computational complexity, which achieves better estimation accuracy.