A model based method for channel estimation and synchronization in a sevearly doppler spread channel is presented. The model parameters are delay, amplitude and doppler (compression/expansion). The parametrs are succesively determined for a number of signal arrivals starting with the strongest and down to the level that is considered significant for the communication. The residual is taken care of by an ordinary non parametric channel estimation. Another novelty is the signal model that is presented. In soft recursive estimation of a message the channel is overestimated when the knowledge of the received symbols is small. To overcome this the phases or amplitudes are constrained to the values that are possible with respect to the modulation format. The results show considerably better convergence in channels with severe doppler shift and varying channel coefficients. The abstract goes here.

We introduce a multi-channel soft input/soft output receiver for underwater communication that performs joint iterative channel estimation, linear equalization, and decoding. The transmitted symbols were encoded using a turbo coded bit-sequence. Our method exploits the gain present in the turbo code through a feed-back of soft information from the decoder to the channel estimator and to the equalizer. The use of several receiving hydrophones makes the method more robust to frequency selective or low-SNR channels, or can improve reception in channels with very large delay spreads. We give examples of such channels measured in the Baltic Sea at two different sea trials during 2005 and 2006. Properties like time variations of the channel impulse response or channel Doppler spread are given, as well as decoding examples using our method. It is shown that the multi-channel version of the receiver achieve error free reception of 4000 coded symbols per second up to 60 km through channels featuring very large delay spreads. The maximum transmitter source level used was 180 dB re 1uPa at 1m at the carrier frequency 12 kHz.