Underwater Acoustic Target Recognition (UATR) remains one of the most challenging tasks in underwater signal processing due to the lack of labeled data acquisition, the impact of the time-space varying intrinsic characteristics, and the interference from other noise sources. Although some deep learning methods have been proven to achieve state-of-the-art accuracy, the accuracy of the recognition task can be improved by designing a Residual Network and optimizing feature extraction. To give a more comprehensive representation of the underwater acoustic signal, we first propose the three-dimensional fusion features along with the data augment strategy of SpecAugment. Afterward, an 18-layer Residual Network (ResNet18), which contains the center loss function with the embedding layer, is designed to train the aggregated features with an adaptable learning rate. The recognition experiments are conducted on the ship-radiated noise dataset from a real environment, and the accuracy results of 94.3% indicate that the proposed method is appropriate for underwater acoustic recognition problems and sufficiently surpasses other classification methods.

Target identification of ship-radiated noise is a crucial area in underwater target recognition. However, there is currently a lack of multi-target ship datasets that accurately represent real-world underwater acoustic conditions. To tackle this issue, we conducted experimental data acquisition, resulting in the release of QiandaoEar22—a comprehensive underwater acoustic multi-target dataset. This dataset encompasses 9 h and 28 min of real-world ship-radiated noise data and 21 h and 58 min of background noise data. To demonstrate the availability of QiandaoEar22, we executed two experimental tasks. The first task focuses on assessing the presence of ship-radiated noise, while the second task involves identifying specific ships within the recognized targets in the multi-ship mixed data. In the latter task, we extracted eight features from the data and employed six deep learning networks for classification, aiming to evaluate and compare the performance of various features and networks. The experimental results reveal that ship-radiated noise can be identified from background noise in over 99% of cases. For the specific identification of individual ships, the optimal recognition accuracy achieves 99.56%. Finally, we found using spectrum and MFCC as feature inputs and DenseNet as classifier can achieve excellent recognition performance. Considering the computational efficiency, CRNN and ECAPA-TDNN are also good choices. Our work not only establishes a benchmark for algorithm evaluation but also inspires the development of innovative methods to enhance underwater acoustic target detection (UATD) and underwater acoustic target recognition (UATR).

Cobalt-rich manganese crusts (CRCs) are important as a potential mineral source that could occur throughout the Pacific on seamounts, ridges, and plateaus. We built a prototype parametric acoustic probe to complete the task of in-situ thickness measurements to estimate the volumetric distribution of deep-sea mineral. The prototype is designed with dual-channels for receiving the primary and secondary signal, which lays a foundation for improving the thickness extraction algorithm. Considering that the signal quality is degraded by the system interference and ambient noise, some improvements to the algorithm are proposed by including the wavelet-based envelope extraction method and the adaptive estimation strategy based on the dual-channel information. Additionally, wavelet regression is applied to reduce the measuring noise assuming that the CRCs have local thickness invariability. The algorithm is suitable for the CRCs with the structure of the multilayers at the top surface and one single layer at the bottom surface. A laboratory experiment is performed to validate the effectiveness of the algorithm. The experiments carried out on the China Ocean 51th voyage in the Western Pacific Ocean on Aug 30, 2018, are described and the data obtained by using the sit-on-bottom stationary measurement are processed to validate the design of the prototype.

Recent studies have shown that pseudo labels can contribute to unsupervised domain adaptation (UDA) for speaker verification. Inspired by the self-training strategies that use an existing classifier to label the unlabeled data for retraining, we propose a cluster-guided UDA framework that labels the target domain data by clustering and combines the labeled source domain data and pseudo-labeled target domain data to train a speaker embedding network. To improve the cluster quality, we train a speaker embedding network dedicated for clustering by minimizing the contrastive center loss. The goal is to reduce the distance between an embedding and its assigned cluster center while enlarging the distance between the embedding and the other cluster centers. Using VoxCeleb2 as the source domain and CN-Celeb1 as the target domain, we demonstrate that the proposed method can achieve an equal error rate (EER) of 8.10% on the CN-Celeb1 evaluation set without using any labels from the target domain. This result outperforms the supervised baseline by 39.6% and is the state-of-the-art UDA performance on this corpus.

The spatial separation of the transmitter and receiver in Bistatic Synthetic Aperture Radar (BiSAR) makes it a promising and useful supplement to a classical Monostatic SAR system (MonoSAR). This paper proposes a novel integrated time and phase synchronization strategy for a multichannel spaceborne-stationary BiSAR system. Firstly, the time synchronization strategy is proposed, which includes Pulse Repetition Frequency (PRF) generation under noisy conditions, multichannel calibration and the alignment of the recorded data with the orbital data. Furthermore, the phase synchronization strategy, which fully considers the deteriorative factors in the BiSAR configuration, is well studied. The contribution of the phase synchronization strategy includes two aspects: it not only compensates the phase error, but also improves the Signal to Noise Ratio (SNR) of the obtained signals. Specifically, all direct signals on different PRF time can be reconstructed with the shift and phase compensation operation using a reference signal. Besides, since the parameters of the reference signal can be estimated only once using the selected practical direct signal and a priori information, the processing complexity is well reduced. Final imaging results with and without compensation for real data are presented to validate the proposed synchronization strategy.