Accurate recognition of radar modulation mode helps to better estimate radar echo parameters, thereby occupying an advantageous position in the radar electronic warfare (EW). However, under low signal-to-noise ratio environments, recent deep-learning-based radar signal recognition methods often perform poorly due to the unsuitable denoising preprocess. In this paper, a denoising-guided disentangled network based on an inception structure is proposed to simultaneously complete the denoising and recognition of radar signals in an end-to-end manner. The pure radar signal representation (PSR) is disentangled from the noise signal representation (NSR) through a feature disentangler and used to learn a radar signal modulation recognizer under low-SNR environments. Signal noise mutual information loss is proposed to enlarge the gap between the PSR and the NSR. Experimental results demonstrate that our method can obtain a recognition accuracy of 98.75% in the −8 dB SNR and 89.25% in the −10 dB environment of 12 modulation formats.

As a promising theory to recover sparse signal from data samples acquired below the Nyquist rate, compressive sensing (CS) has been drawing pervasive interest in the past decade. In this paper, we explore the compressive sensing potentials for the near-field multiple acoustic-source localization. A novel localization scheme is designed by introducing the optimization of the measurement matrix to enforce the restricted isometry property (RIP) and maximize the signal-to-noise ratio (SNR). Monte Carlos simulations have been carried out to demonstrate the effectiveness of our proposed new scheme. Compared to other existing localization techniques, our scheme exhibits superior performances.

The increasing penetration of renewable energy sources and the widespread use of power electronic devices have introduced complex power quality disturbances (PQDs), which are essentially caused by the superposition and coupling of multiple single disturbances. Accurate identification and classification of complex PQDs are crucial prerequisites for effective power quality assessment and control. However, existing classification methods often neglect the correlations among disturbance category labels, limiting their applicability under multi-disturbance conditions. To address these limitations, this paper proposes a Transformer-based multi-label classification framework that integrates label semantics fusion for complex PQDs. Specifically, a multi-scale dilated residual convolution module is employed to extract discriminative features across multiple temporal scales; the label semantics fusion module then embeds disturbance label semantics into the signal features; finally, a Transformer encoder is introduced along the category dimension to explicitly model inter-label dependencies, thereby enhancing multi-label classification accuracy. Furthermore, extensive validation on both the standard IEEE PES dataset and a self-developed physical experimental platform confirms the reliability and robust generalization capability of the proposed approach across heterogeneous measurement scenarios.

In this letter, we first present the design and implementation of a customized aquatic robotic node. Based on the robotic vision platform, we propose an improved contamination detection model based on lightweight network YOLOX-Nano, in which attention fusion schemes are integrated to strengthen the features beneficial to contaminant detection, so as to improve the performance of the model. The proposed network has only 0.91 MB in size, but achieves the mean average precision (mAP@0.5:0.95) of 35.2% on a self-made dataset, ensuring the detection precision and speed. Besides, the proposed network has been applied to water environment monitoring, running at a detection speed of 9.5 frames per second (FPS) on the aquatic node. Experimental results demonstrate that it is effective and feasible for aquatic contaminant detection using the proposed robotic vision system.