Research On Deterministic Blind Channel Identification Algorithms

In the digital wireless communication system, for the multipath propagation and the impact of band-limited channel, it is hard to avoid intersymbol interference at the receiver. In order to compensate for distortion, channel identification is necessary. Blind channel identification technique without or with little known symbols has been widesspread concerned. Blind channel identification technique based on second-order statistics(SOS) has become one of the focuses in blind channel identification field on account of its finite sample convergence property and low complexity. Based on a great defense project, this paper mainly researches on the deterministic blind channel identification algorithms based on second-order statistics which are suitable for practical applications. The main content and work are outlined as follows:1. The SIMO channel model is incroduced firstly. The general identifiability conditions, such as channel, source and data, are summarized in the blind channel identification algorithms based on SOS. Meanwhile, the relationship between the channel zero distribution and the identifiability of deterministic blind channel identification algorithms based on SOS is studied with related theory analysis and simulation verification.2. Targeting at the problem that the effect of the deterministic blind channel identification algorithms based on SOS strongly depends on the channel order, a modified algorithm for channel order estimation via sample sorting is proposed. The algorithm constructs the two-dimensional space of successive sample rank pairs and utilizes the particular graph structure that is under-estimation to estimate the channel order. The modified algorithm has a better estimation performance than the original one at low SNR. Then a channel order estimation algorithm based on identification and equalization is proposed. The algorithm constructs a convex identification cost function and proposes a weighted least square equalization criterion with relative theoretical analysis. Joint with two convex cost function mentioned, the channel order is estimated when the algorithm gets its global optimal solution. Simulation results show that the proposed algorithm has a better estimation performace than other existing order estimation algorithms under the condition of different channels. It is also stable and reliable.3. In order to improve the robustness to channel order error of the deterministic blind channel identification algorithms based on SOS, this paper firstly makes a theoretical analysis on the relationship between channel zero distribution and the channel order that is overestimated. It is found that the extra estimated channel common zeros introduced by overestimating are gathered around the unit circle. According to this theorem, a blind channel identification algorithm based on the distribution of channel zero is proposed. The algorithm is sample and wild adaptability. Joint with this theorem and a modified CR algorithm which has low computational complexity, a blind channel identification algorithm that is robust to order overestimation using the FFT method is proposed. The algorithm derives channel response in the frequency domain so as to improve the identification performace when the number of sample data is small. Simulation results show that the proposed algorithm has a good robustness to channel order error.4. Targeting at the problem that blind channel identification algorithms cannot estimate the channel that has common zero and are sensitive to channel order error, a semi-blind channel identification algorithm based on singular value decomposition is proposed. The algorithm decomposes the channel matrix into two matrices in the form of product based on singular value decomposition and uses the receive data and some known symbols to estimate the channel response. Simulation results have verified the feasibility and effectiveness of the proposed algorithm. Then a cross-relation-based semi-blind channel identification algorithm is proposed. The algorithm builds a linear system of equations based on the orthogonal relationship between the correlation matrix made by output data and channel vector. It also utilizes some known symbols based on the modified least square criterion to build other equations. The closed form solution of channel response is derived by using least-square method. The proposed algorithm is steady and has a good identification performance. It also has a good robustness to channel noise and channel order error.