Single channel blind signal separation of time-frequency overlapped signals based on independent component analysis

In domain of blind signal processing (BSS), it is often the case that many sources are received by one antenna. Extracting one or several of them separately becomes a significant research task. A method of single-channel and time-frequency overlapped multi-component signals separation is proposed in this paper, which uses the technique of changing single-input multi-output (SIMO) system into virtual multi-input multi-output (MIMO) model by a dynamical embedding framework. And then FastICA (Independent Component Analysis) algorithm is adopted to perform on the embedding matrix decomposing the single-channel time-frequency overlapped recording into its underlying independent components (ICs), which can be regards as a convenient expansion basis. After that identity components of interest relevant to the original signals, and project back onto the measurement space in isolation. The simulation results demonstrate the effectiveness of the single-channel multi-component separation method presented in this paper. At last, the analysis of this method is shown in detail.