基于最大序列相关性的Turbo码交织器识别

针对现有算法在Turbo码交织器识别中存在低信噪比适应性差，且识别性能随交织长度增加而急剧恶化的缺点，提出了基于最大校正序列相关的识别算法。该算法首先利用已识别的交织位置序列对每一帧交织位置上的信息序列进行估计，通过遍历可能的交织位置，并作互相关运算，当遍历位置为交织映射关系时，该位置上的数据序列与估计序列具有最大的相似度，从而完成交织位置识别；然后充分利用这些交织位置上的序列，分别估计出校正数据序列，将校正的序列再与原始序列叠加，完成码元校正，直到所有的交织关系识别完成。所提算法直接利用了截获的软判决信息，同时能够实现码元校正，这就克服了以往算法的两个缺点。仿真结果表明，在信噪比为-1 dB，交织长度为1024时，所提出算法仅仅需要1000数据帧，就能达到100%的识别率，与以往算法相比，性能提升2~3 dB，同时完成一次可靠识别所需的数据量仅需以往算法的1/4。

Blind recognition of interleaver for Turbo codes based on maximum sequence correlation

The existing algorithms in the recognition of interleaver for Turbo codes have poor adaptability to low SNR and deteriorate sharply with the increase of interleaving length. To address these disadvantages, a new algorithm based on the maximum correction sequences correlation is proposed. First, the algorithm estimates the information sequences at each interleaved position by using the identified interleaved positions sequences. The possible interleaved positions are traversed and the cross-correlation operation is performed. When the traversing position is an interleaved mapping relation, the data sequences at that position have the greatest similarity with the estimated sequences, thus completing recognition. Then, by making full use of the sequences on the interleaved positions, the corrected data sequence is estimated and the corrected sequence is superimposed with the original sequence to complete symbol correction until all the interleaved relations are identified. The proposed algorithm utilizes the intercepted soft decision information directly and achieves symbol correction at the same time, overcoming two shortcomings of the previous algorithms. Simulation results show that when the SNR is -1 dB and interleaving length is 1024, the recognition rate can reach 100% with 1000 blocks of data frames. Compared with the previous algorithms, the performance of the proposed algorithm is improved by 2 to 3 dB, and the amount of data needed to complete a reliable identification is reduced to less than 1/4 of the original data.