Under the condition of non-cooperative wireless communication, many signals always overlap in time–frequencyfield, therefore, the signal separation and reconstruction of the received mixed signals is of great significance for the subsequent information processing. A new blind separation strategy is proposed to solve the blind separation problem in non-cooperative communication under general underdetermined conditions. Firstly, based on a new double-constrained single source points (SSP) detection criterion, a fuzzy mean clustering underdetermined blind identification (UBI) algorithm is proposed which got the high precision estimation of the mixing matrix. Then a singular value membership matching underdetermined source recovery (SVMMUSR) algorithm with dynamic k sparse component analysis (kSCA) assumption is present. The singular value decomposition (SVD) method is applied to detect the membership of every sample data point with the subspace so as to obtain the optimal k-dimensional subspace matching with each data point. Subspace projection method is then used to achieve the accurate recovery of the signal for unknown k sparse conditions. Compared with other conventional methods, the simulation results indicate that the estimation performance and blind separation performance of the proposed method is better.

在非合作无线通信的情况下，许多信号在时频域中总是重叠，因此，接收信号的分离和重构对于后续的信息处理具有重要意义。提出了一种新的盲分离策略，以解决一般不确定条件下非合作通信中的盲分离问题。首先，基于一种新的双约束单源点（SSP）检测准则，提出了一种模糊均值聚类不确定盲识别（UBI）算法，该算法对混合矩阵进行了高精度估计。然后采用动态k稀疏分量分析（k）的奇异值隶属度匹配不确定源恢复（SVMMUSR）算法SCA）假设存在。奇异值分解（SVD）方法用于检测每个样本数据点与子空间的隶属关系，以获得与每个数据点匹配的最优k维子空间。然后使用子空间投影方法在未知的k个稀疏条件下实现信号的准确恢复。仿真结果表明，与其他常规方法相比，该方法的估计性能和盲分离性能更好。