关键词: 欠定盲分离, AR模型, 功率谱估计, 扩展子空间, FastICA

Abstract: Underdetermined blind source separation is studied in this paper.In order to make use of time-frequency characteristic of signal，a new method of variational estimation on signal frequency by using AR model power spectrum estimation，designing band-pass filter to gain approximatively source signals and underdetermined mixed matrix，and constructing completed observation signals by expanded subspace primer vector is proposed.Thus original blind source separation problem is transformed from underdetermined to completed situation.Finally blind source separation is achieved by FastICA method.The simulation experiment results show that the proposed method is feasible and effective，which provides new way for study on underdetermined blind source sepatation.

Key words: underdetermined blind source separation, AR model, power spectrum estimation, expanded subspace, FastICA