Unpaired image-to-image translation aims to preserve the semantics of the input image while mimicking the style of target domains without paired data. However, existing methods often suffer from ...semantic distortions if the source and target domains have large mismatched semantics distributions. To address semantic distortions in translation outputs without paired supervision, we propose a Margin Adaptive Contrastive Learning Network (MACL-Net) that drives contrastive learning as a local semantic descriptor while using a pre-trained Vision Transformer (ViT) as a global semantic descriptor to learn domain-invariant features in the translation process. Specifically, we design a novel margin adaptive contrastive loss to enforce intra-class compactness and inter-class discrepancy. Besides, to better retain the semantic structure of the translated image and improve its fidelity, we use Discrete Wavelet Transform (DWT) to supplement the low-frequency and high-frequency information of the input image into the generator, and effectively fuse the feedforward features and inversed frequency information through a novel normalization scheme, Feature-Frequency Transformation Normalization (FFTN). In terms of experimental results, MACL-Net effectively reduces semantic distortions and generates translation outputs that outperform state-of-the-art techniques both quantitatively and qualitatively.
•We propose MACL-Net to reduce image artifacts and objecthallucinations.•We propose a novel Margin Adaptive Contrastive Loss (MACL).•We design an Adaptive Frequency Fusion (AFF) module.•We design a novel normalization scheme Feature-Frequency Transformation Normalization (FFTN).
•A novel model entitled CDTFAFN is constructed for multi-sensory information fusion.•An improved method named ICQ-NSGT is presented for time-frequency transformation.•A feature fusion unit termed ...TFA-FFU is designed for enhancing feature learning.•A machinery intelligent fault diagnosis framework based on CDTFAFN is proposed.•Case studies prove the effectiveness and superiority of the proposed method.
When the machinery device operates abnormally, it is not sufficient for fault detection only via extracting fault features from a single sensor due to the latent fault information may be scattered across multiple sensors. Multi-sensory fusion techniques with deep learning framework have attracted increasing attention from researchers due to the exploiting and integration of fault information between multiple sensors. Nevertheless, there are two remaining shortcomings in most existing multi-sensory fusion technologies. (1) Most existing fusion methods merely concentrate on conducting multi-sensory information fusion from time-domain or frequency-domain to achieve fault diagnosis, which are often unsatisfactory in the face of strong noise environments. (2) The collaborative fusion between several vibration sensors is generally considered in the past works, whereas the complementary information fusion between multi-sensory vibro-acoustic heterogeneous data are rarely studied. To address these deficiencies, this paper proposes a novel coarse-to-fine dual-scale time-frequency attention fusion network (CDTFAFN) for machinery fault diagnosis, which not only adequately considers the complementary information fusion of vibro-acoustic signal, but also has robust feature learning capabilities in a noisy scenario. Firstly, the signal-to-image encoding unit (SIEU) containing the improved constant-Q non-stationary Gabor transform (ICQ-NSGT) is introduced to convert the collected raw vibro-acoustic heterogeneous signal into time-frequency representation (TFR) and achieve the coarse-grained feature fusion. Secondly, the time-frequency attention feature fusion unit (TFA-FFU) is designed to concurrently learn the fine-grained features at two scales from the low-level fused features which are meaningful for fault diagnosis. Finally, the coarse-to-fine features are sequentially concatenated and fed into softmax classifier to preferably promote the network learning performance and automatically implement fault classification. The performance of the proposed approach is validated against those state-of-the-art results on two groups of multi-sensory vibro-acoustic data in different experimental platforms. Experiment results show that the proposed method with the diagnosis accuracy of 99 % above outperforms other several representative fusion technologies (i.e., 2MNet, MFF-GBFD, MSCNN-BiLSTM, MFAN-VAF, 1D-CNN-VAF, MI-CNN-TFT and TFFN-VAF) in the raw noise-free addition scenario. Moreover, the average testing accuracy of the proposed method can still reach 97 % above in the noisy scenarios with Gaussian white noises, which shows its competitive superiority and strong robustness against noises in machinery fault diagnosis. According to the five ensemble macro-average performance evaluation metrics (i.e., accuracy, precision, sensitivity, specificity and F1-score) and the receiver operator characteristic (ROC) analysis, our findings also emphasize the superiority of applying our method for machinery fault diagnosis under the colored noises compared with other fusion technologies (i.e., 2MNet, MFF-GBFD, MSCNN-BiLSTM, MFAN-VAF, 1D-CNN-VAF, MI-CNN-TFT and TFFN-VAF) reported in this paper.
Residence time difference (RTD) fluxgate sensor is a potential device to measure the DC or low-frequency magnetic field in the time domain. Nevertheless, jitter noise and magnetic noise severely ...affect the detection result. A novel post-processing algorithm for jitter noise reduction of RTD fluxgate output strategy based on the single-frequency time difference (SFTD) method is proposed in this study to boost the performance of the RTD system. This algorithm extracts the signal that has a fixed frequency and preserves its time-domain information via a time⁻frequency transformation method. Thereby, the single-frequency signal without jitter noise, which still contains the ambient field information in its time difference, is yielded. Consequently, compared with the traditional comparator RTD method (CRTD), the stability of the RTD estimation (in other words, the signal-to-noise ratio of residence time difference) has been significantly boosted with sensitivity of 4.3 μs/nT. Furthermore, the experimental results reveal that the RTD fluxgate is comparable to harmonic fluxgate sensors, in terms of noise floor.
The identification of patterns and underlying characteristics of natural or engineering time-varying phenomena poses a challenging task, especially in the scope of simulation models and accompanying ...stochastic models. Because of their complex nature, time-varying processes such as wind speed, seismic ground motion, or vibrations of machinery in the presence of degradation oftentimes lack a closed-form description of their underlying Evolutionary Power Spectral Density (EPSD) function. To overcome this issue, a wide range of measurements exist for these types of processes. This opens up the path to a data-driven stochastic representation of EPSD functions. Rather than solely relying on time–frequency transform methods like the familiar short-time Fourier transform or wavelet transform for EPSD estimation, a probabilistic representation of the EPSD can provide valuable insights into the epistemic uncertainty associated with these processes. To address this problem, the evolutionary EPSD function is relaxed based on multiple similar data to account for these uncertainties and to provide a realistic representation of the time data in the time–frequency domain. This results is the so-called Relaxed Evolutionary Power Spectral Density (REPSD) function, which serves as a modular probabilistic representation of the time–frequency content of stochastic signals. For this purpose, truncated normal distributions and kernel density estimates are used to determine a probability density function for each time–frequency component. The REPSD function enables the sampling of individual EPSD functions, facilitating their direct application to the simulation model through stochastic simulation techniques like Monte Carlo simulation or other advanced methods. Even though the accuracy is highly dependant on the data available and the time–frequency transformation method used, the REPSD representation offers a stochastic representation of characteristics used to describe stochastic signals and can reduce epistemic uncertainty during the modelling of such time-varying processes. The method is illustrated by numerical examples involving the analysis of dynamic behaviour under random loads. The results show that the method can be successfully employed to account for uncertainties in the estimation of the EPSD function and represent the accuracy of the time–frequency transformation used.
•Probabilistic representation of environmental & engineering processes in time–frequency domain.•Accounting for epistemic uncertainties inherent in environmental processes and stochastic signals.•Utilisation of parametric and non-parametric PDF functions for EPSD functions.•Applicable to both separable and non-separable EPSD functions.•Robust simulation results in structural response analysis accounting for uncertainties.
This article presents a new generalized direct synthesis method to design multiband bandpass filters (MBPFs). Based on the proposed frequency transformation technique, it can be used to synthesize an ...arbitrary number of bandpass filters starting from a low-pass prototype filter. This synthesis allows determining analytically all the resonant angular frequencies and slope parameters of all the bandpass filters from the low and high cut-off angular frequencies used as initial specifications whatever the number of bands. To validate our method, triple-, quad-, and quint-band third-order Chebyshev bandpass filters are designed and implemented in microstrip technologies. Very good agreements were achieved between simulation responses and measurements.
Image fusion plays a crucial role in enhancing the quality and accuracy of semantic segmentation, which is essential for autonomous driving systems. By merging information from multiple imaging ...sensors or modalities, such as infrared and visible images, image fusion enriches the data and improves the perception capabilities of autonomous vehicles. However, current fusion methodologies often cannot balance model complexity, inference efficiency, and fusion accuracy simultaneously, making them difficult to implement in resource-constrained environments. In response to this, this paper presents a lightweight fusion network based on frequency transformation and deep learning techniques, leveraging wavelet transformation to fuse infrared and visible images. Concisely, the fusion model decomposes input images into different frequency sub-bands using wavelet transforms. It then efficiently fuses the multi-scale feature representations in the frequency domains with a specially designed fusion loss. Compared to traditional fusion approaches, our method not only achieves a better balance between subjective fusion quality and downstream vision tasks but also significantly improves model inference efficiency, paving the way for real-time autonomous driving systems. Extensive experiments on public datasets show that our method can achieve state-of-the-art performance while satisfying parameter efficiency in the context of image fusion and semantic segmentation tasks. Concisely, our approach is nearly 100× faster while using a model 6000× smaller in size compared to SegMIF.
Driver's cognitive workload has an important impact on driving safety. This paper carries out an on-road experiment to analyse the impact from three innovative aspects: significance analysis of ...electroencephalogram (EEG) under different cognitive workloads, distribution of EEG maps with different frequency signals and influence of different cognitive workloads on driving safety based on EEG. First, the EEG signals are processed and four frequencies of delta, theta, alpha and beta are obtained. Then, the time-frequency transform and power spectral density calculation are carried out by short-time Fourier to study the correlation of each frequency signal of different workload states, as well as the distribution pattern of the EEG topographic map. Finally, the time and space energy and phase changes in each cognitive task event are studied through event-related spectral perturbation and inter-trial coherence. Results show the difference between left and right brains, as well as the resource occupancy trends of the monitor, perception, visual and auditory channels in different driving conditions. Results also demonstrate that the increase in cognitive workloads will directly affect driving safety. Changes in cognitive workload have different effects on brain signals, and this paper can provide a theoretical basis for improving driving safety under different cognitive workloads. Mastering the EEG characteristics of signals can provide more targeted supervision and safety warnings for the driver.
Semisupervised change detection (CD) methods have garnered increasing attention due to their capacity to alleviate the dependency of fully-supervised methods on a large number of pixel-level labels. ...These methods predominantly leverage generative adversarial network architecture and consistency regularization technology. However, they encounter challenges associated with background noise from cross-temporal images. In this article, we propose a novel multilevel consistency-regularization-based semisupervised CD approach that incorporates Fourier-based frequency transformation and a reliable pseudolabel selection scheme. Specifically, we replace the low-frequency spectrum of one temporal image with a frequency domain transformation derived from the corresponding image in the same bitemporal remote sensing image pair, enhancing the model's capability to discern meaningful changes amidst background noise, thereby contributing to more robust CD. Furthermore, excessively high pseudolabel thresholds in consistency regularization methods may result in the underutilization of valuable unlabeled data. To address this issue, we design a straightforward sigmoid-like function to dynamically adjust the selection threshold for the reliable pseudolabel selection scheme. This strategy takes into consideration the learning status throughout the entire training process, ensuring more effective utilization of unlabeled information. We demonstrate significant performance improvements across three widely-used public datasets, namely, LEVIR-CD, WHU-CD, and CDD. Notably, on the three datasets with only 1% labeled data, our method achieved an <inline-formula><tex-math notation="LaTeX">\text{IoU}^{c}</tex-math></inline-formula> of 71.29%, 63.90%, and 51.00%, outperforming existing state-of-the-art methods by 2.84%, 1.21%, and 0.98%, respectively. These results robustly substantiate the effectiveness of our approach, showcasing its potential in scenarios where labeled data is limited.
In the above paper, S. C. Dutta Roy proposed design principles of single and multi-band L-type lumped matching networks. It is a fundamental work to understand the multi-band operation of lumped ...matching networks. However, some typos in the paper may lead to the erroneous design of dual-band and triple-band matching networks. In this letter, we will provide the correct design equations.
A systematic design method for high-order dual-band bandpass frequency selective surfaces (FSSs) with a low profile is derived from classical filter theory and presented here. To complement the ...design procedure, a multilayer double-slot resonator unit cell topology is proposed for realizing dual-band operations. For simplicity, the resonators are made to work for only a single polarization. To design the FSS, first, a classical dual-band bandpass filter circuit is designed by performing successive frequency transformations on a lowpass prototype. The filter is then transformed into a form resembling the equivalent circuit of the proposed multilayer FSS structure. Finally, the transformed filter is mapped to a set of FSS geometrical parameters. The method presents very few inherent limitations to realizing a diverse range of filter responses. The resulting designs lend themselves to fabrication since very few layers of metallization are required. Two FSSs with third-order passbands at 4 and 7 GHz but different passband characteristics are designed and verified numerically. One of the designs is fabricated and experimentally verified. The overall thickness of the designs is <inline-formula> <tex-math notation="LaTeX">0.08\lambda _{l} </tex-math></inline-formula> where <inline-formula> <tex-math notation="LaTeX">\lambda _{l} </tex-math></inline-formula> is the free-space wavelength at 4 GHz. The unit cell size is approximately <inline-formula> <tex-math notation="LaTeX">\lambda _{l}/8 </tex-math></inline-formula>.