盲信号恢复是从退化的信号中识别出真实的信号,由于其有广泛的应用前景,因而引起了众多学者的研究兴趣。非负矩阵分解作为一种强大的特征提取方法,已经在许多领域得到了很好的应用。本文把非负矩阵分解理论运用到盲信号的分离问题中,提出了一种基于定点非负矩阵分解(FPNMF)的盲信号恢复方法。该方法特点是借鉴盲源分离理论直接从观察信号中辨识原信号。实验中对大量模拟数据进行了盲处理和分析,结果表明本文提出的方法具有较强的有效性和鲁棒性。 Blind signal recovery is to identify the true signal from the degraded signal, which has aroused the interest of many scholars due to its wide application prospect. As a powerful feature extraction method, non-negative matrix factorization has been well applied in many fields. In this paper, we apply the theory of non-negative matrix factorization to the separation of blind signals and propose a blind signal recovery method based on fixed-point non-negative matrix factorization (FPNMF). The method is characterized by the blind source separation theory is used to directly recognize the original signal from the observed signal. A large number of simulation data were blindly processed and analyzed in the experiment. The results show that the proposed method is effective and robust.