A blind signal separation method for single-channel electromagnetic surveillance system

Abstract

In this paper, a blind signal separation (BSS) methodology for simultaneously received multisystem frequency-overlapped signals in a single-channel (SC) electromagnetic surveillance system is proposed using fast independent component analysis (FastICA) in a dynamical embedding (DE) framework. Firstly, an appropriate DE matrix is constructed out of a series of delay vectors from the SC recording. The lag-time and the dimensional of embedding matrix setting principal are introduced in details. Next, multiple independent components (ICs) are calculated by decomposing the embedding matrix through FastICA algorithm, and ICs can be regarded as a convenient expansion basis of the original signals. Then, these ICs are projected back into the measurement space. After that, these projected ICs are classified and used for recovering the sources of interest based on their independent nature and their power density spectrum. Numerical simulation results obtained in evaluating the proposed methodology's performance confirmed the effectiveness of the proposed algorithm.

A MLE-based blind signal separation method for time–frequency overlapped signal using neural network

“…ICA and its variants only require that the source signals are independent to each other and have been successfully applied in all kinds of signal separation, like speech signal [66], biomedical signal-e.g., electroencephalographic (EEG) and magnetoencephalographic (MEG) [67], and so on. What's more, ICA also can be used to undetermined signal separation problem under certain condition that the under-determined observation matrix can be transformed into an observation matrix whose rank is no less than source signal number [18]. For signals with sparsity, sparse component analysis (SCA) is another popular method, and it has been successfully separate image mixture separation [23][24][25], speech signal [26,27], biological signal [28,29] and so on.…”

“…Novey et al [15] proposed a complex fast independent component analysis (c-FastICA) algorithm to solve the ICA problems with complex-valued data. FastICA algorithm has been successfully applied in different fields, such as electroencephalography (EEG) processing [16,17], single-channel digital communication signal separation [18], modern power systems [19] and joint radar and communication signal separation [20]. Additionally, some researchers finished the implication of FastICA algorithm, like Shyu et al [21] implemented the FastICA algorithm in a field-programmable gate array (FPGA), with the ability of real-time sequential mixed signals processing by the proposed pipelined FastICA architecture.…”