In this paper, a quadratic convolution neural network (QCNN) using both audio and vibration signals is utilized for bearing fault diagnosis. Specifically, to make use of multi-modal information for bearing fault diagnosis, the audio and vibration signals are first fused together using a 1 × 1 convolution. Then, a quadratic convolution neural network is applied for the fusion feature extraction. Finally, a decision module is designed for fault classification. The proposed method utilizes the complementary information of audio and vibration signals, and is insensitive to noise. The experimental results show that the accuracy of the proposed method can achieve high accuracies for both single and multiple bearing fault diagnosis in the noisy situations. Moreover, the combination of two-modal data helps improve the performance under all conditions.

Clarity of audio signal collected by voyage data recorders (VDR) is of great significance to reliable investigation of voyage accidents or vessal situation recovery. However, the VHF audio signal in VDR is often buried in various noises, resulting in risky understanding of audio content. To address this issue, a cross-representation loss guided complex convolutional network (CRGCCN) is proposed. It consists of a complex encoding, a complex decoding and a complex Conformer modules. In this work, absolute errors in frequency domain ( L _Log–PCM ) and relative errors in time domain ( L _Si–SNRi ) are integrated together in the cross-representation loss function, resulting in reasonable network parameters. In the proposed loss function, L _Log–PCM contributes to reduce the absolute errors, and L _Si–SNRi accounts for improving the robustness of the proposed network to different signal-to-noise ratios. Experimental results show that the proposed model achieves best performance on both synthesized and real-world VHF audio datasets compared to several state-of-the-art methods.

Very High Frequency (VHF) communication instruments are widely applied in vessels for transmitting both regular and emergency information. However, VHF communication speech often deteriorates severely due to complex transmission conditions. Most existing speech enhancement (SE) models ignore the correlation of different time-frequency (T-F) units in the frequency domain or temporal samples in the time domain. Additionally, existing loss functions cannot effectively model spectral details with less intensities. As a result, speech partial loss and residual noise problems are serious for low signal-to-noise ratio (SNR) VHF communication speech. To address these problems, a speech enhancement network is proposed in this work. More specifically, a coarse-refine strategy is introduced to better disentangle speech and noise. In the coarse stage, the noise is preliminarily removed based on the Encoder-Conformer-Decoder backbone and the consistency module. In the refine stage, the speech and noise partials are further disentangled based on consistency and contrastive learning modules. The consistency module makes the enhanced speech consistent with clean speech at both fine-grained and global levels, and contrastive learning module can better distinguish clean speech from noise in hidden feature space. Experimental results demonstrate that the proposed network obtains 28.0% higher PESQ score and 3.8% higher STOI score than SOTA network on synthesized dataset. Also, it achieves 4.5% higher MOS score than SOTA network on real-world dataset.