Linear Prediction-Based Detection of Serially Concatenated DQPSK in SIMO-OFDM

An approach for the linear prediction-based detection of serially concatenated turbo coded differential quaternary phase shift keyed signals, is presented for SIMO-OFDM systems. The proposed method exploits the high degree of correlation in the channel frequency response, when the channel impulse response is much smaller than the FFT length. A prediction filter is used to estimate the channel frequency response. At the receiver, the inner decoder operates on a supertrellis with just $$S_{\mathrm {ST}}=S_{\mathrm{E}}\times 2^{P-1}$$SST=SE×2P-1 states, where the complexity reduction is achieved by using the concept of isometry (here $$S_{\mathrm {E}}$$SE denotes the number of states in the encoder trellis and P denotes the prediction order). A reduced complexity version based on per-survivor processing for the proposed receiver is also given. Simulation results for the proposed method is compared to the case where the channel is estimated using pilots in the frequency domain. Though the BER performance of the channel estimation approach using pilots is much better than the proposed method, its throughput is much lower, since the pilots have to be transmitted for every Orthogonal Frequency Division Multiplexing (OFDM) frame. However, in the proposed approach, the pilots are required only for the first OFDM frame for the purpose of estimating the statistical properties of the channel frequency response and noise. For the rest of the frames, pilots are not required since, the channel and noise statistics are assumed to be unchanged. Therefore, the proposed method is throughput efficient.