A novel single‐channel blind source separation method is studied. This method not only addresses the problem that the traditional blind source separation method depends on the number of sensors but also achieves the purpose of effectively identifying the modal parameters. Firstly, the time varying filter for empirical mode decomposition (TVF‐EMD) method is used to decompose the one‐dimensional observation signal into intrinsic mode functions (IMFs) with different scale characteristics. Then, the similarity measure of the probability density function between each IMF and the one‐dimensional observation is calculated. According to the threshold, several sensitive IMFs are selected and a new observation is constructed with random partial reconstruction to form a two‐dimensional matrix. Finally, the reconstructed signals are separated by the improved sparse component analysis (SCA) method based on energy detection in the frequency domain. The simulation results demonstrate that the proposed method can separate not only simulated vibration signal, but also damping sinusoidal signal effectively from the single‐channel while the source number is unknown. And the parameters of natural frequencies and damping ratios of modal responses can be accurately identified in the test of the measured cantilever beam hammering.

Blind modulation recognition (BMR) involves identifying the modulation scheme of intercepted signals, an essential component of terrestrial radio management. While deep learning (DL) has advanced BMR research in terrestrial contexts, 6 G multi-beam mobile satellite (MMS) systems present many challenges. One pivotal hurdle is the reliance on perfect channel-state information (CSI) for signal equalization in terrestrial MIMO BMR literature. In MMS systems, the vast satellite-ground distance can lead to outdated CSI, making traditional equalization techniques less effective. Addressing this, we introduce a viable BMR algorithm that leverages blind and semi- blind channel equalization to overcome challenges like inter-beam interference (IBI), limited CSI, and Shadowed-Rician (SR) fading. By transforming the MMS system into an equivalent MIMO satellite system, we enable enhanced MIMO channel equalization to counteract IBI and SR attenuations. Our proposed locality-globality convolutional-transformer deep neural network (LG-CTDNet) merges the strengths of CNNs in local feature extraction with Transformers in global feature discernment. A decision fusion strategy is then incorporated to utilize cooperative diversity across multiple eavesdroppers. Numerical tests validate the effectiveness of our blind and semi- blind equalization techniques, emphasizing their superiority over existing DL-based methods in satellite modulation signal identification even in conditions with limited or absent CSI.

Modulation recognition (MR) plays a pivotal role due to its application in the spectrum sensing of 5G industrial cognitive communications and radio interference detection at the physical layer of the Internet of Things (IoT). Previous works have mainly focused on simulated fourth-generation (4G) multicarrier systems and ideal radio frequency (RF) scenarios. To bridge the gap between practice and theory, we propose a viable MR algorithm on all-physical testbeds, with nonlinear impairments of 28 GHz active phased arrays (APA). Specifically, our testbed is built on the Rohde&Schwarz (R&S) vector signal generation R&S-SMBV100B and spectrum analyzer R&S-FSW 67 GHz. To extract salient modulation patterns, we develop a physical-informed scattering transform (SCT) MR network (SCTMR-Net). With SCT modules, SCTMR-Net produces the translation-invariant and deformation-stable representations of 5-G signals by wavelet convolution, nonlinear modulus and low-pass filters. Extensive experiments on real-world measurement verify the viability of SCTMR-Net for high robustness to APA impairments.

In order to explore the effective statistical information of observed signals to stack into a tensor, while reduce the computational complexity and speed up the convergence, a new tensor-based underdetermined blind identification method of instantaneous mixture is proposed in this brief. Firstly, the autocovariance matrix of each observed sub-block is calculated to construct the symmetric third-order tensor according to segmentation scheme. Secondly, the original tensor is reduced into a low-rank kernel tensor by truncated multi-linear singular value decomposition (MLSVD). Finally, an enhanced line search (ELS) technique is applied to speed up the convergence of alternating least squares (ALS) to identify the mixing matrix. The simulation results indicate that the presented approach provides a superior estimated performance compared to the existing methods in both real and complex scenarios.