Time-frequency Image Research Articles

GNSS blanket jamming classification algorithm based on spatial attention mechanism and residual shrinkage neural network

In the context of an increasingly complex electromagnetic environment, satellite navigation systems have become highly susceptible to jamming. Detecting and classifying jamming has thus become crucial for taking effective anti-jamming measures. This paper addresses the issue that the classification accuracy of blanket jamming declines drastically in low jamming-to-noise ratio (JNR) scenarios. To tackle this challenge, a novel algorithm is proposed that combines the spatial attention mechanism with a residual shrinkage neural network (RSN-SA) to classify ten types of blanket jamming, ranging from single jamming to convolutional compound jamming. Specifically, the proposed algorithm first employs the Fourier Synchrosqueezed Transform to extract time-frequency (TF) domain features from the original jamming signal, generating corresponding TF images. Then, the RSN-SA is employed to identify and classify these images effectively while minimizing the impact of noise-related features. This allows the main parts of the TF images to be focused on, resulting in higher recognition accuracy. Simulation results demonstrate that RSN-SA achieves close to 100% accuracy for six single blanket jamming signals. Moreover, compared with the other five algorithms, RSN-SA effectively enhances the classification accuracy of convolutional compound jamming signals in low JNR scenarios and improves the recognition stability in high JNR scenarios. Overall, the proposed algorithm provides a promising solution for classifying blanket jamming in satellite navigation systems with high accuracy and robustness.

Machine Fault Diagnosis through Vibration Analysis: Continuous Wavelet Transform with Complex Morlet Wavelet and Time–Frequency RGB Image Recognition via Convolutional Neural Network

In pursuit of advancing fault diagnosis in electromechanical systems, this research focusses on vibration analysis through innovative techniques. The study unfolds in a structured manner, beginning with an introduction that situates the research question in a broader context, emphasising the critical role of fault diagnosis. Subsequently, the methods section offers a concise summary of the primary techniques employed, highlighting the utilisation of short-time Fourier transform (STFT) and continuous wavelet transform (CWT) for extracting time–frequency components from the signal. The results section succinctly summarises the main findings of the article, showcasing the results of features extraction by CWT and subsequently utilising a convolutional neural network (CNN) for fault diagnosis. The proposed method, named CWTx6-CNN, was compared with the STFTx6-CNN method of the previous stage of the investigation. Visual insights into the time–frequency characteristics of the inertial measurement unit (IMU) data are presented for various operational classes, offering a clear representation of fault-related features. Finally, the conclusion section underscores the advantages of the suggested method, particularly the concentration of single-frequency components for enhanced fault representation. The research demonstrates commendable classification performance, highlighting the efficiency of the suggested approach in real-time scenarios of fault analysis in less than 50 ms. Calculation by CWT with a complex Morlet wavelet of six time–frequency images and combining them into a single colour image took less than 35 ms. In this study, interpretability techniques have been employed to address the imperative need for transparency in intricate neural network models, particularly in the context of the case presented. Notably, techniques such as Grad-CAM (gradient-weighted class activation mapping), occlusion, and LIME (locally interpretable model-agnostic explanation) have proven instrumental in elucidating the inner workings of the model. Through a comparative analysis of the proposed CWTx6-CNN method and the reference STFTx6-CNN method, the application of interpretability techniques, including Grad-CAM, occlusion, and LIME, has played a pivotal role in revealing the distinctive spectral representations of these methodologies.

Deep Anomaly Detection with Attention (DADA): A Novel Approach for Identifying Multipath Interference in Radar Signals

Multipath interference in radar signals caused by sea, ground, and other environments poses significant challenges to the target detection, tracking, and classification capabilities of radar systems. Existing methods for radar signal identification require labeled samples and focus mainly on the classification of normal signals. However, in practice, anomalous samples (multipath interference signals) may be scarce and highly imbalanced (i.e., mostly normal samples). To address this problem, we propose a deep anomaly detection with attention (DADA) for semisupervised detection of multipath radar signals. The method transforms radar signals into time–frequency images and is trained exclusively on normal samples. The autoencoder architecture is extended with a feature extractor network to capture latent sample features. CBAM attention is introduced to improve feature extraction. By learning the distribution of normal samples in high-dimensional image space and low-dimensional feature space, a two-dimensional feature space representing normal samples is constructed. A one-class SVM then learns the boundary of normal samples for anomaly detection. Extensive experiments on radar signal datasets validate the effectiveness of the proposed approach.

A fault diagnosis method for nuclear power plants rotating machinery based on deep learning under imbalanced samples

Rotating machinery is a critical equipment widely used in nuclear power plants (NPPs). Effective fault diagnosis technology can provide reliable operation and maintenance support for rotating machinery. Data-driven intelligent fault diagnosis technology has attracted much attention in recent years. However, in practical situations, the available fault data is limited, and the imbalanced samples will make the intelligent fault diagnosis model prone to problems such as poor generalization performance and low diagnostic accuracy. Therefore, a fault diagnosis method based on deep learning is proposed to alleviate the impact of imbalanced samples on fault diagnosis of rotating machinery. First, the adaptive synthetic sampling (ADASYN) approach is used to synthesize the multi-channel vibration data to expand the unbalanced samples. Subsequently, ensemble empirical mode decomposition (EEMD) and continuous wavelet transform (CWT) are used to convert the vibration data into time–frequency images to highlight the fault features of the samples. Then, the deep residual neural network is constructed to extract the features of mixed samples and implement fault diagnosis. Fault simulation experiments are carried out based on motors and bearings, and various imbalance degree datasets are formed to provide data support for verifying the effectiveness of the proposed method. The proposed method can achieve good diagnostic performance under various degrees of imbalanced samples. In addition, the comparison results with other methods show that the proposed method has the best comprehensive performance, demonstrating the potential application value in the fault diagnosis of NPPs rotating machinery.

Radar Signal Recognition Based on CNN With a Hybrid Attention Mechanism and Skip Feature Aggregation

Radar signal recognition is of great significance for threat analysis and intelligence support in radar electronic warfare. However it remains a challenging task due to (1) the variety of radar signal modulation types, (2)the increasingly saturated electromagnetic environment and the more serious influence of noise interference, and (3)the intrinsic real-time requirement of this application. In this paper, a novel automatic modulation recognition algorithm based on the joint architecture of convolution neural network (CNN) and the support vector machine (SVM) is proposed to comprehensively address these challenges. In particular, this algorithm first employs time-frequency analysis to better show the pulse representation in the time-frequency domain. Then a well-designed DCNN equipped with a hybrid attention (HA) mechanism and skip feature aggregation (SFA) is used to automatically learn the feature representations of distinctiveness from time frequency images. HA is designed to quickly capture the important information of local parts in sight and enhance the ability of the backbone network to mine high-level features. SFA is applied to selectively aggregate features among channels for self-monitoring, which improves the expression and generalization ability. Sixteen types of radar signals are used to verify the feasibility and effectiveness of this algorithm. The experimental evaluation shows that the proposed method not only achieve competitive accuracy with a better timeliness than state-of-the-art methods, but also achieves effective identification of most of the signal modulation types.

Hand Gesture Classification through Time Frequency Images with One-to-One, All-to-One, and Cross-Validation Approaches.

Surface electromyographic (sEMG) signals are a non-invasive method for acquiring signals that play a fundamental role in the monitoring of prosthetic devices by providing information about human motor functions. This leads to the need for accurate classification of sEMG signals, despite variations in signal stationarity, the presence of sensor noise, differences between the muscles involved, and the peculiarities of each patient. This study focuses on the classification of hand grip postures using sEMG signals acquired from amputee patients. Special emphasis is placed on the use of the time-frequency domain for feature extraction, using the spectral analysis of the reduced-time Fourier transform (STFT).To carry out this task, a classification model based on a convolutional neural network (CNN) is used. The classification method is adjusted, trained, and evaluated through three experiments. The first, called "One to One", yields accuracy percentages of 90.84%, 91.05%, and 91.13% for spectrograms of 32x32, 64x64, and 128x128 in size, respectively. In the second validation, called "All by One", an accuracy of 62.28% is achieved for spectrograms of 32x32 pixels. Finally, in the last K-fold cross-validation validation, an average accuracy of 86.73%, 86.77%, and 87.97% is obtained for spectrograms of 32x32, 64x64, and 128x128 in size, respectively.

Modeling and Prediction of Nonstationary Ground Motions as Time–Frequency Images

Proper modeling of nonstationary characteristics of ground motions is critical in estimating the seismic performance of structures. The importance of ground motion nonstationarity is being increasingly recognized in seismic design codes and provisions. This paper develops a new method for predicting seismic ground motions in the joint time–frequency domain based on recorded ground motions. We treat wavelet energy maps of acceleration records as images and extract essential patterns from them using principal component analysis. These patterns are then linked to the seismic source, path, and site variables using general regression neural network. The resulting model predicts an image representing the expected evolutionary power spectrum conditioned on the input variables. An example application is presented using records from the Next Generation Attenuation of Ground Motions Database of the Pacific Earthquake Engineering Center. As opposed to conventional ground motion prediction models, the proposed approach retains the time domain characteristics of ground motions. The results show that the proposed model can predict significant patterns in the seismic energy distribution, acceleration time series, and 5% damped elastic spectral accelerations.

Bearing fault diagnosis with parallel CNN and LSTM.

Intelligent diagnosis of bearing faults is fundamental to machinery automation and their intelligent operation. Deep learning-based analysis of bearing vibration data has emerged as one research mainstream for fault diagnosis. To enhance the quality of feature extraction from bearing vibration signals and the robustness of the model, we construct a fault diagnostic model based on convolutional neural network (CNN) and long short-term memory (LSTM) parallel network to extract their temporal and spatial features from two perspectives. First, via resampling, vibration signal is split into equal-sized slices which are then converted into time-frequency images by continuous wavelet transform (CWT). Second, LSTM extracts the time-correlation features of 1D signals as one path, and 2D-CNN extracts the local frequency distribution features of time-frequency images as another path. Third, 1D-CNN further extracts integrated features from the fusion features yielded by former parallel paths. Finally, these categories are calculated through the softmax function. According to experimental results, the proposed model has satisfactory diagnostic accuracy and robustness in different contexts on two different datasets.

CVAN: A Novel Sleep Staging Method Via Cross-View Alignment Network.

Sleep staging is imperative for evaluating sleep quality and diagnosing sleep disorders. Extant sleep staging methods with fusing multiple data-views of physiological signals have achieved promising results. However, they remain neglectful of the relationship among different data-views at different feature scales with view position-alignment. To address this, we propose a novel cross-view alignment network, termed cVAN, utilising scale-aware attention for sleep stages classification. Specifically, cVAN principally incorporates two sub-networks of a residual- like network which learn spectral information from time-frequency images and a transformer- like network which learns corresponding temporal information. The prime advantage of cVAN is to adaptively align the learned feature scales among the different data-views of physiological signals with a scale-aware attention by reorganizing feature maps. Extensive experiments on three public sleep datasets demonstrate that cVAN can achieve a new state-of-the-art result, which is superior to existing counterparts. The source code for cVAN is accessible at the URL (https://github.com/Fibonaccirabbit/cVAN).

Mechanical Fault Diagnosis With Noisy Multisource Signals via Unified Pinball Loss Intuitionistic Fuzzy Support Tensor Machine

In this paper, a challenging and significant intelligent fault diagnosis task is investigated, in which multisource sensor signals with intense noise and outlier disturbances are jointly analyzed. Such a scenario has hardly been considered in industrial research. To this end, we develop a novel tensor-based nonlinear classifier called unified pinball loss intuitionistic fuzzy support tensor machine (UPIFSTM), which can successfully solve the above tasks and improve the performance of fault diagnosis in practical applications. First, the noisy multisource signals are converted into time-frequency images and reconstructed into tensor samples to mine the time domain, frequency domain features and coupled structure information in the spatial domain. Next, we design two nonlinear forms of nonmembership functions in the tensor space, and obtain an intuitionistic fuzzy score for each training sample to enhance the robustness of the model. Subsequently, pinball loss function is introduced to better handle noise sensitivity and resampling instability problems. Note that, we employ the tensor robust principal component analysis method to accurately recover the low-rank tensors corrupted by sparse noise from the original tensors. Finally, two numerical examples are presented to verify the feasibility and validity of the proposed method.

Associations Between Sleep Spindle Metrics, Age, Education and Executive Function in Young Adult and Middle-Aged Patients with Obstructive Sleep Apnea.

This study aimed to investigate the association between sleep spindle metrics and executive function in individuals with obstructive sleep apnea (OSA). Furthermore, we examined the association of age and education on executive function. A total of 230 (40.90 ± 8.83 years, F/M = 45/185) participants were enrolled. Overnight electroencephalogram (C3-M2) recording detected sleep spindles by a novel U-Net-type neural network that integrates temporal information with time-frequency images. Sleep spindle metrics, including frequency (Hz), overall density (events/min), fast density (events/min), slow density (events/min), duration (sec) and amplitude (µV), were measured. Executive function was assessed using standardized neuropsychological tests. Associations between sleep spindle metrics, executive function, and demographic factors were analyzed using multivariate linear regression. In fully adjusted linear regression models, higher overall sleep spindle density (TMT-A, B=-1.279, p=0.009; TMT-B, B=-1.813, p=0.008), fast sleep spindle density (TMT-A, B=-1.542, p=0.048; TMT-B, B=-2.187, p=0.036) and slow sleep spindle density (TMT-A, B=-1.731, p=0.037; TMT-B, B=-2.449, p=0.034) were associated with better executive function. And the sleep spindle duration both during N2 sleep time (TMT-A, B=-13.932, p=0.027; TMT-B, B=-19.001, p=0.034) and N3 sleep time (TMT-B, B=-29.916, p=0.009; Stroop-incongruous, B=-21.303, p=0.035) was independently associated with better executive function in this population. Additionally, age and education were found to be highly associated with executive function. Specific sleep spindle metrics, higher overall density, fast density and slow density during N2 sleep time, and longer duration during N2 and N3 sleep time, are independent and sensitive indicators of better executive function in young adult and middle-aged patients with OSA. Further research is needed to explore the underlying mechanisms and clinical implications of these findings.

A Cross-Scale Transformer and Triple-View Attention Based Domain-Rectified Transfer Learning for EEG Classification in RSVP Tasks.

Rapid serial visual presentation (RSVP)-based brain-computer interface (BCI) is a promising target detection technique by using electroencephalogram (EEG) signals. However, existing deep learning approaches seldom considered dependencies of multi-scale temporal features and discriminative multi-view spectral features simultaneously, which limits the representation learning ability of the model and undermine the EEG classification performance. In addition, recent transfer learning-based methods generally failed to obtain transferable cross-subject invariant representations and commonly ignore the individual-specific information, leading to the poor cross-subject transfer performance. In response to these limitations, we propose a cross-scale Transformer and triple-view attention based domain-rectified transfer learning (CST-TVA-DRTL) for the RSVP classification. Specially, we first develop a cross-scale Transformer (CST) to extract multi-scale temporal features and exploit the dependencies of different scales features. Then, a triple-view attention (TVA) is designed to capture spectral features from triple views of multi-channel time-frequency images. Finally, a domain-rectified transfer learning (DRTL) framework is proposed to simultaneously obtain transferable domain-invariant representations and untransferable domain-specific representations, then utilize domain-specific information to rectify domain-invariant representations to adapt to target data. Experimental results on two public RSVP datasets suggests that our CST-TVA-DRTL outperforms the state-of-the-art methods in the RSVP classification task. The source code of our model is publicly available in https://github.com/ljbuaa/CST_TVA_DRTL.

Adversarial subdomain adaptation method based on multi-scale features for bearing fault diagnosis

Due to the variable working environment of bearings, the collected data often follow different probability distributions. It is hard to directly use the trained models to identify the bearing fault with different operating conditions. In addition, it is a high cost to label the samples for every work condition. To solve these problems, a multi-scale adversarial subdomain adaptation bearing fault diagnosis method is proposed, which is based on Continuous Wavelet Transform (CWT) and our constructed Multi-scale Adversarial SubDomain Adaptation Network (MASDAN). Firstly, to extract the features of non-stationary signals with different frequency bands, CWT is used to convert continuous vibration signals into two-dimensional time-frequency images. Secondly, to enhance the correlation of the features across frequency bands, a multi-scale ConvNeXt is proposed, which adds a multi-scale module based on ConvNeXt to obtain features with different scales. Finally, to reduce the distribution discrepancy between the source domain and the target domain and to avoid feature confusion in different domains, a domain adaptive alignment network introducing domain information is constructed. Two modules are included: the Domain Alignment Classification Network Module (DACNM) based on Multi-kernel Local Maximum Mean Discrepancy (MK-LMMD) and the Domain Adversarial Network Module (DANM) based on domain discrimination. Thus, the MASDAN consists of the multi-scale ConvNeXt module, DACNM, and DANM, which can realize the multi-scale feature extraction and adaptively align the fault features at different domains. The experimental results on the Qingdao University of Technology (QUT) bearing dataset and the Case Western Reserve University (CWRU) bearing dataset demonstrate the proposed method can effectively diagnose bearing faults under different operating conditions.

Leveraging Transfer Learning for Data Augmentation in Fault Diagnosis of Imbalanced Time-Frequency Images

Leveraging Transfer Learning for Data Augmentation in Fault Diagnosis of Imbalanced Time-Frequency Images

Research on the Application of Image Feature Extraction in Mechanical Structure Recognition and Fault Diagnosis

Abstract With the rapid development of modern industry and science and technology, in recent years the fault diagnosis method based on image processing has become a research hotspot in the field of mechanical fault diagnosis. In this paper, image characteristics are extracted from multiple aspects such as image texture, color, shape, etc. A grayscale symbiotic matrix image feature extraction method is proposed. On this basis, the algorithm for extracting gray symbiotic matrix time-frequency image features is designed. At the same time, the algorithm and parameters of mechanical structure identification are optimized to identify and diagnose mechanical faults. The results show that the grayscale symbiotic matrix time-frequency image feature extraction algorithm is able to accurately diagnose the wear-type faults, overwork-type faults, and short-circuit-type fault behavior of the mechanical equipment. All of them are able to obtain more than 80% accuracy, and all of them are able to reach 99.99% accurate detection of mechanical faults, which proves the effectiveness of the method of this research.

Sleep Stage Classification Via Multi-View Based Self-Supervised Contrastive Learning of EEG.

Self-supervised learning (SSL) is a challenging task in sleep stage classification (SSC) that is capable of mining valuable representations from unlabeled data. However, traditional SSL methods typically focus on single-view learning and do not fully exploit the interactions among information across multiple views. In this study, we focused on a multi-domain view of the same EEG signal and developed a self-supervised multi-view representation learning framework via time series and time-frequency contrasting (MV-TTFC). In the MV-TTFC framework, we built-in a cross-domain view contrastive learning prediction task to establish connections between the temporal view and time-frequency (TF) view, thereby enhancing the information exchange between multiple views. In addition, to improve the quality of the TF view inputs, we introduced an enhanced multisynchrosqueezing transform, which can create high energy concentration TF image views to compensate for the inaccurate representations in traditional TF processing techniques. Finally, integrating temporal, TF, and fusion space contrastive learning effectively captured the latent features in EEG signals. We evaluated MV-TTFC based on two real-world SSC datasets (SleepEDF-78 and SHHS) and compared it with baseline methods in downstream tasks. Our method exhibited state-of-the-art performance, achieving accuracies of 78.64% and 81.45% with SleepEDF-78 and SHHS, respectively, and macro F1-scores of 70.39% with SleepEDF-78 and 70.47% with SHHS.

Advancements in condition monitoring and fault diagnosis of rotating machinery: A comprehensive review of image-based intelligent techniques for induction motors

Recently, condition monitoring (CM) and fault detection and diagnosis (FDD) techniques for rotating machinery (RM) have witnessed substantial advancements in recent decades, driven by the increasing demand for enhanced reliability, efficiency, and safety in industrial operations. CM of valuable and high-cost machinery is crucial for performance tracking, reducing maintenance costs, enhancing efficiency and reliability, and minimizing mechanical failures. While various FDD methods for RM have been developed, these predominantly focus on signal processing diagnostics techniques encompassing time, frequency, and time-frequency domains, intelligent diagnostics, image processing, data fusion, data mining, and expert systems. However, there is a noticeable knowledge gap regarding the specific review of image-based CM and FDD. The objective of this research is to address the aforementioned gap in the literature by conducting a comprehensive review of image-based intelligent techniques for CM and fault FDD specifically applied to induction motors (IMs). The focus of the study is to explore the utilization of image-based methods in the context of IMs, providing a thorough examination of the existing literature, methodologies, and applications. Furthermore, the integration of image-based techniques in CM and FDD holds promise for enhanced accuracy, as visual information can provide valuable insights into the physical condition and structural integrity of the IMs, thereby facilitating early FDD and proactive maintenance strategies. The review encompasses the three main faults associated with IMs, namely bearing faults, stator faults, and rotor faults. Furthermore, a thorough assessment is conducted to analyze the benefits and drawbacks associated with each approach, thereby enabling an evaluation of the efficacy of image-based intelligent techniques in the context of CM and FDD. Finally, the paper concludes by highlighting key issues and suggesting potential avenues for future research.

The research of interference recognition method in multi-nodes cooperative frequency-hopping communication based on time–frequency image analysis and deep learning

Interference recognition has a significant and irreplaceable importance in keeping battlefield communication reliability and securing information safety. However, the existing recognition algorithms have a common disadvantage, that the model scale is normally very huge. In order to satisfy the demand of decreasing mobile device weight and limiting the memory space, this article designed a lightweight neural network based on the idea of structure reparameterization. This network will take the time–frequency image of the interference signal as the input of the convolutional neural network (CNN), and import the efficient channel attention module, efficiently extract the feature of the interference signals, thus benefiting the interference recognition. The simulation experiment indicates that the network has a high accuracy rate in interference recognition in the circumstances when the quantity of parameters is not huge. At the same time, in order to solve the problem of low recognition accuracy due to the limit of information, this article designed an interference recognition algorithm based on multi-nodes collaboration technology. Considering jamming-to-noise ratio (JNR) could differ for each node, this article set a node select module, that could select the nodes that contain a higher result of JNR as the input of the fusion algorithm when each JNR is highly different, which could adjust abnormal influence from each node. The simulation experiment indicates that, the interference recognition algorithm based on multi-nodes collaboration technology could efficiently increase interference recognize accuracy, and reduce the influence from low JNR nodes.

Deep Learning-Based Visual Complexity Analysis of Electroencephalography Time-Frequency Images: Can It Localize the Epileptogenic Zone in the Brain?

In drug-resistant epilepsy, a visual inspection of intracranial electroencephalography (iEEG) signals is often needed to localize the epileptogenic zone (EZ) and guide neurosurgery. The visual assessment of iEEG time-frequency (TF) images is an alternative to signal inspection, but subtle variations may escape the human eye. Here, we propose a deep learning-based metric of visual complexity to interpret TF images extracted from iEEG data and aim to assess its ability to identify the EZ in the brain. We analyzed interictal iEEG data from 1928 contacts recorded from 20 children with drug-resistant epilepsy who became seizure-free after neurosurgery. We localized each iEEG contact in the MRI, created TF images (1–70 Hz) for each contact, and used a pre-trained VGG16 network to measure their visual complexity by extracting unsupervised activation energy (UAE) from 13 convolutional layers. We identified points of interest in the brain using the UAE values via patient- and layer-specific thresholds (based on extreme value distribution) and using a support vector machine classifier. Results show that contacts inside the seizure onset zone exhibit lower UAE than outside, with larger differences in deep layers (L10, L12, and L13: p < 0.001). Furthermore, the points of interest identified using the support vector machine, localized the EZ with 7 mm accuracy. In conclusion, we presented a pre-surgical computerized tool that facilitates the EZ localization in the patient’s MRI without requiring long-term iEEG inspection.

Dynamics simulation-driven fault diagnosis of rolling bearings using security transfer support matrix machine

Transfer fault diagnosis holds paramount significances in safeguarding the reliability and safety of rolling bearings. However, the current studies require massive experiment or field data in the source domain. Besides, sparse data (only single or several samples for each fault) and strong-noise data in the target domain are prone to cause the negative transfer problem. Importantly, most transfer fault diagnosis methods conduct fault classification in vector space, so the learning ability is weak for matrix-form fault features such as 2D time-frequency images. To address these issues, this paper proposes a novel dynamics simulation-driven fault diagnosis framework with security transfer support matrix machine (STSMM). In this framework, a high-fidelity bearing dynamic model is designed to generate sufficient source-domain data, which greatly reduces the acquisition cost of real data. A new matrix-form transfer learning model, i.e., STSMM, is proposed to effectively utilize the structural information contained in matrix data and achieve the simulation-to-real transfer of fault knowledge. Besides, a security transfer strategy is embedded in STSMM to improve the cross-domain diagnosis performance and theoretically avoid the negative transfer problem caused by sparse data and strong-noise data in the target domain. Experiment results demonstrate the effectiveness and superiority of the proposed framework.