中文摘要：

目前盲分离研究已有算法众多,但有关数字信号或有限字符集的盲分离研究尚不多见,而带源个数估计的此类盲分离算法更鲜有涉及。针对这类问题,该文提出了一种新颖的BPSK数字信号的盲分离算法,首先由接收到的观测信号的特征,在无噪和有噪情况下分别给出了估计源信号的数目算法;然后再利用观测信号之间的关系估计出混叠矩阵,并在算法中给出了证明。通过估计的混叠矩阵即可恢复得到分离信号,此时得到的分离信号与源信号或者顺序发生了交换,或者产生了符号之间的差别,但并不影响盲分离。最后的仿真结果显示了该文提出的算法在估计混叠矩阵以及最后恢复源信号上都是非常成功,也表明了此算法的可行性和优异性能。

英文摘要：

At present, there exist a lot of algorithms of blind separation, among which there are few algorithms focusing on blind separation of digital signals and estimating source number. To solve the problem, this paper proposes a novel algorithm to solve the blind separation problem of BPSK signals. First, according to characteristics of observations, the algorithm to estimate the source number is given in noise circumstance and no noise circumstance. Second, mixing matrix is gotten by using the relation of observations, which is also proved feasible. Finally, the source signals are recovered by permutations and sign changes of their rows, which are allowed in blind separation. It is well shown that the algorithms are excellent and feasible to estimate the mixed matrix and recover source in the last simulation.