Research On Channel Estimation Based On Decision-Directed Technique

ABSTRACT:Orthogonal-frequency-division-multiplexing (OFDM) has been widely accepted in modern wireless communication systems, such as Long-Term Evolution (LTE) based4G wireless communication systems, due to its robustness against frequency selectivity in wireless channels. In OFDM systems, a frequency selective channel is con-verted into a collection of flat fading channels, which can be compensated by simply using a one-tap equalizer. Hence, OFDM can greatly simplify equalizer design while enabling rather high data rates. As a result, OFDM has been standardized as a key physical-layer technique in many commercial systems.For coherent detection, channel estimation is necessary for OFDM receiver design. Traditional channel estimation techniques are based on pilot symbols. To improve the channel estimation accuracy, decision-directed (DD) techniques can be adopted where decided symbols are fed back to the channel estimator. Assuming the decided symbols are correct, it is equivalent to increasing the pilot density, and thus leads to improved channel estimation quality. Traditional DD channel estimation uses only the feedback of hard decision. In this case, DD manner is unique, that is, feeding the hard decided symbols directly back to the channel estimator. With the development of Turbo receivers, soft decision has been widely applied, resulting in significant performance improvement. Rather than a unique hard decision result, soft decision generates a set of a-posteriori probabilities (AP) conditioned on the received signal for each point in the constellation. Correspondingly, for a DA channel estimator, it obtains the set of APs rather than a unique hard decision. Since the APs for each constellation point is now available, there are different ways for DA channel estimation. As a result, the DA manner is not unique any more, but with many possibilities. Therefore, it needs to address the optimal data-aided channel estimation problem.For low and medium mobility environment, time variation of channels over one OFDM symbol is small and thus can be omitted. Therefore, channels are assumed to be constant in traditional channel estimation techniques. However, as the increase of termi-nal mobility, the Doppler frequency also increases. In this situation, channel varies fast and channel time variation over one OFDM symbol cannot be omitted any more. If the assumption of constant channel over one OFDM symbol is still used, it will leads to great estimation error. Therefore, channel variation over one OFDM symbol should be tak-en into account to improve the channel estimation quality in fast time-varying channels. Traditional channel estimation approaches are not suitable for this case, and thus it needs to address the problem of channel estimation in fast time-varying channels.A Maximum A-posteriori Probability (MAP) based iterative technique is proposed, to address the optimal DA channel estimation for single transmit antenna single OFDM symbol case. The obtained results show that harmonic average should be employed to generate the soft data symbols required by the channel estimator. Although there is no analytical solution for the derived MAP equation, it can be solved iteratively with fixed-point iteration (FPI) approach. Further results show that the proposed method is closely connected with the expectation-maximization (EM) algorithm, and this connection can be used to explain the convergence of the proposed method. Meanwhile, it is also proved that the proposed method can converge within one step when the signal-to-noise ratio (SNR) is large enough. Simulation results show that the proposed method can approach the Cramer-Rao bound (CRB) when the SNR is large.The MAP base iterative technique is then extended to the case of multiple transmit antenna and multiple OFDM symbols. The analysis shows that when there are multiple transmit antennas, a harmonic-average based soft symbols on expected transmit anten-na while an arithmetic average based soft symbols on the interfering transmit antennas. Meanwhile, considering multiple OFDM symbols continually transmitted in the time-domain, time-variation and time-correlation properties can be exploited to improve the channel estimation performance. Similar the single transmit antenna single OFDM sym-bol case, the iterative channel estimation is derived based on the MAP criterion, and can be iteratively solved using FPI approach. Convergence can be also guaranteed for the pro-posed method, meanwhile, the one-step convergence property still holds when the SNR is large enough.Channel estimation for OFDM systems over fast time-varying channels is also in-vestigated. Time-varying channel is represented using a basis expansion model (BEM). For the BEM, the basis functions are determined, and it only needs to estimate the BEM coefficients (BEMC). In this paper, by taking account into the time correlation among consecutive OFDM blocks, the optimum BEMC estimation is derived in terms of the minimum mean-square-error (MMSE) criterion. Using the proposed approach, the esti-mates of BEMCs for current OFDM block can be derived through a linear combination of adjacent BEMCs and there is no need to insert pilot symbols into each OFDM blocks. Practical receiver designs are also considered in this paper, including the finite-impulse response (FIR) filter based implementation and an iterative receiver structure.