Marine ambient noise can negatively affect underwater acoustic target (UWAT) recognition. Previous related studies have focused on the suppression of high-frequency noise. However, marine ambient noise in the frequency domain is concentrated at low frequencies, overlapping with the signal components of UWATs. Low-rank (LR) matrix approximation is an effective class of denoising methods, but its direct application on UWAT signals tends to result in the loss of weak signal components. To better suppress low-frequency noise, we propose a denoising method based on mode decomposition and LR matrix approximation. This method first decomposes the UWAT signal into a series of modes using improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), which disperses the signal components into different modes thus emphasizing weak signal components. Subsequently, an adaptive dual judgment method based on amplitude-aware permutation entropy (AAPE), cosine similarity (CS), and K-means++ is applied to all modes to identify the signal modes and then discard the noise modes for initial denoising. Finally, an improved OptShrink algorithm which can adaptively choose the rank by clustering and shrink singular values is proposed to extract the LR signal matrix for each signal mode and further suppress the low-frequency noise in the signal modes. Experimental results on the ShipsEar dataset show that our method can effectively suppress low-frequency noise. More importantly, the difference between UWATs with different labels is also enhanced after employing our proposed method.

Many studies have used various time-frequency feature extraction methods to convert ship-radiated noise into three-dimensional (3D) data suitable for computer vision (CV) models, which have shown good results in public datasets. However, traditional feature engineering (FE) has been enhanced to interface matching–feature engineering (IM-FE). This approach requires considerable effort in feature design, larger sample duration, or a higher upper limit of frequency. In this context, this paper proposes a one-dimensional network design for underwater acoustic target recognition (UATR-ND1D), only combined with fast Fourier transform (FFT), which can effectively alleviate the problem of IM-FE. This method is abbreviated as FFT-UATR-ND1D. FFT-UATR-ND1D was applied to the design of a one-dimensional network, named ResNet1D. Experiments were conducted on two mainstream datasets, using ResNet1D in 4320 and 360 tests, respectively. The lightweight model ResNet1D_S, with only 0.17 M parameters and 3.4 M floating point operations (FLOPs), achieved average accuracies were 97.2% and 95.20%. The larger model, ResNet1D_B, with 2.1 M parameters and 5.0 M FLOPs, both reached optimal accuracies, 98.81% and 98.42%, respectively. Compared to existing methods, those with similar parameter sizes performed 3–5% worse than the methods proposed in this paper. Additionally, methods achieving similar recognition rates require more parameters of 1 to 2 orders of magnitude and FLOPs.

The attention mechanism improves underwater acoustic target recognition (UATR) by suppressing irrelevant features. However, due to the uncertainty and scarcity of underwater acoustic target (UWAT) signals, complicated deterministic attention modules increase the risk of model overfitting, resulting in limited improvement or even degradation in the performance of UATR. This letter proposes a Bayesian Hybrid Attention Module (BHAM) that enhances UATR based on time–frequency (T–F) features. BHAM models attention weights as random variables following Beta and Dirichlet distributions to capture uncertainty of UWAT signals and mitigate overfitting, while strengthening T–F feature representation via Bayesian channel attention and Bayesian T–F attention. By learning attention distributions in a Bayesian manner, BHAM effectively models complex dependencies in UWAT signals. Experiments on the DeepShip dataset demonstrate that BHAM alleviates overfitting and generalizes well across different network backbones.

Underwater acoustic target (UWAT) signal denoising is challenging due to the complexity of the marine environment. Most current denoising methods suffer from poor low- signal-to-noise ratio (SNR) adaptability, limited generalizability, and high vulnerability to target signal component loss. To address these issues, we propose a dual-path (DP) flow-enhanced noise-aware (NA) variational autoencoder (DFNA-VAE) for UWAT signal denoising. DFNA-VAE employs a time–frequency (T–F) attention-based encoder to learn the T–F feature differences between noise and UWAT signals. Subsequently, the NA probabilistic latent Space of DFNA-VAE models the latent representations with a probabilistic framework, and disentangles the latent representations into signal latent variables and noise latent variables, which are then input into two decoders to generate noise and target signals, respectively. This design enhances the model’s generalization capability and robustness. Meanwhile, a DP inverse autoregressive flow is proposed to enhance the representational capacity of the latent space. Finally, we derivate the variational evidence lower bound of DFNA-VAE to guide the optimization of the model in variational Bayesian inference framework. Experimental results demonstrate that DFNA-VAE can achieve good denoising performance on UWAT signals with low SNRs and exhibits good generalization capability. More significantly, DFNA-VAE can reduce the loss of signal components during denoising, thereby enhancing the interclass differences of UWATs, suitable for the preprocessing of UWAT recognition.

This paper proposes a broadband beamforming method for underwater target radiated-noise measurement using a hybrid vector–pressure array that combines a vector nested line subarray with a pressure uniform circular array. A hybrid measurement model is established, and a unified steering vector is derived by jointly accounting for subarray geometries and sensor responses. Simulations are conducted using a plane-wave far-field baseline and two three-path waveguide environments with varying multipath strengths. Performance is quantified by continuous-spectrum reconstruction errors over 250–2000 Hz, excluding tonal components. Results demonstrate that the vector-assisted scheme substantially mitigates ambiguity-induced leakage and yields markedly lower reconstruction errors in waveguide conditions, maintaining overall MAE/RMSE at the 5–7 dB level, lower than that of conventional and CBW about 12-15dB. This demonstrates improved robustness to multipath propagation and enhanced port–starboard ambiguity suppression for wideband radiated-noise reconstruction.