In recent years, the underwater acoustic sensor network (UASN) is emerging as an effective means for marine data collection. However, due to the limited bandwidth of acoustic channel, the limited constrained energy su...
详细信息
ISBN:
(纸本)9798350386288;9798350386271
In recent years, the underwater acoustic sensor network (UASN) is emerging as an effective means for marine data collection. However, due to the limited bandwidth of acoustic channel, the limited constrained energy supply of underwater sensors and data redundancy, it is impossible to deliver all the raw data generated by underwater sensors. Hence, to compress time-redundant data at a sensor node is of great significance for data collection from underwater sensors. Moreover, the underwater acoustic transmission link is unreliable, the issue of data collection with certain packet error resilience should be resolved. In this paper, a packet-loss-and-error-resilient data collection method based on convolutional auto-encoder (CAE) is proposed to collect the time-series data from underwater sensors, named as the PLER-CAE data collection method. In the proposed method, the auto-encoder is used to reduce data redundancy of the time-series data. The encoder deployed at the collection end compresses the data for transmission through the underwater channel, and the decoder deployed at the receiving end reconstructs the original data. Moreover, the proposed method complements the retransmission mechanism and the error correction technique in order to enhance data reconstruction quality. Numerical results show that the proposed PLER-CAE data collection method is effective.
This letter focuses on the cross-corpus speech emotion recognition (SER) task, in which the training and testing speech signals in cross-corpus SER belong to different speech corpora. Existing algorithms are incapable...
详细信息
This letter focuses on the cross-corpus speech emotion recognition (SER) task, in which the training and testing speech signals in cross-corpus SER belong to different speech corpora. Existing algorithms are incapable of effectively extracting common sentiment information between different corpora to facilitate knowledge transfer. To address this challenging problem, a novel convolutional auto-encoder and adversarial domain adaptation (CAEADA) framework for cross-corpus SER is proposed. The framework first constructs a one-dimensional convolutional auto-encoder (1D-CAE) for feature processing, which can explore the correlation among adjacent one-dimensional statistic features and the feature representation can be enhanced by the architecture based on encoder-decoder-style. Subsequently the adversarial domain adaptation (ADA) module alleviates the feature distributions discrepancy between the source and target domains by confusing domain discriminator, and specifically employs maximum mean discrepancy (MMD) to better accomplish feature transformation. To evaluate the proposed CAEADA, extensive experiments were conducted on EmoDB, eNTERFACE, and CASIA speech corpora, and the results show that the proposed method outperformed other approaches.
Fault diagnosis is an important subfield of prognostic and health management (PHM). Intelligent fault diagnosis based on deep learning is the most popular data-driven method of the present. However, current researches...
详细信息
Fault diagnosis is an important subfield of prognostic and health management (PHM). Intelligent fault diagnosis based on deep learning is the most popular data-driven method of the present. However, current researches are prone to ignoring the strong noisy backgroundin real working conditions and cannot achieve excellent performance in actual application. As we all know, noise reduction and feature extraction are two vital aspects in mechanicalfault diagnosis. In this article, an intelligent diagnostic model based onimproved stacked convolutional auto-encoder (ISCAE) and parallel attention-based convolutional blocks (PACB) is proposed. ISCAE-based module is constructed to reduce the noise of raw signals and then PACB-based module can synchronouslyextract local spatial feature and global feature *** equalize the role of above-mentioned two modules which are serial in the proposed model, two modules are trained and optimized synchronously to simultaneously adjust the neural network weights. The capability and effectiveness of the model are evaluated using a dataset collected from real operating environment of main reducer. The comparative analysisresults show that the ISCAE-PACB-based model can reach the accuracy of 98.95% and is superior to existing models.
This study presents a novel bridge anomaly detection approach that employs the reconstruction error and structural similarity of an unsupervised convolutional auto-encoder. The presence of structural damage in a bridg...
详细信息
This study presents a novel bridge anomaly detection approach that employs the reconstruction error and structural similarity of an unsupervised convolutional auto-encoder. The presence of structural damage in a bridge typically results in changes in its vibration signals, and thus, the use of these signals for structural damage detection (SDD) has been widely investigated, with many methods relying on supervised learning. However, such existing SDD methods based on the supervised learning require prior knowledge of the damage states and cannot process monitoring data in real-time, thereby limiting their application to in-service bridges. To address this challenge, the authors propose the use of a convolutional auto-encoder as the reconstruction algorithm for real-time vibration signals. The auto-encoder is trained using normal signals and then used to reconstruct new inputs (either normal or abnormal). Two damage indicators (reconstruction error and structural similarity) are then calculated based on the reconstruction results and clustered to detect abnormal signals. The proposed approach was applied to the detection of various abnormalities in the old ADA Bridge, the results were 100% accurate, and about a 10% increase in accuracy was observed when compared to other control experiments. These results demonstrate the effectiveness of the proposed approach, with the auto-encoder achieving excellent reconstruction results for normal signals and clear discrepancies for abnormal signals. The proposed method was also validated on a cable-stayed bridge and an arch bridge, demonstrating its wide applicability in bridge anomaly detection.
Secure transfer of digital images is an important issue in modern image communications. The complexity of encryption algorithms and their key space is a guarantee for security and robustness issues, and the algorithms...
详细信息
Secure transfer of digital images is an important issue in modern image communications. The complexity of encryption algorithms and their key space is a guarantee for security and robustness issues, and the algorithms are competing in this regard while considering time complexity constraints. In this paper, it is tried to increase the encryption complexity and unpredictability of the encryption scheme using different phases of chaos game representation (GCR), logistic map diffusion, and convolutional auto-encoder-based image representation. In the proposed scheme, the original image's pixels are first turned into a binary sequence, after which the CGR algorithm is applied to create a sequence of coordinates in the CGR space. These coordinates are used as the first key, and the pixels of the original image are aligned with this key. Additionally, the second key is generated using the chaotic logistic map to increase the complexity of the key. After that, a randomly weighted convolutional auto-encoder is applied, and the encrypted image resulting from the first two stages is fed to the network for simultaneous diffusion and compression for further complexity increment. The applied convolutional auto-encoder does not require training, and it offers a significant amount of key space, which helps robustness against various attacks. Resulting from the experiments, the average entropy of the encrypted images is relatively high, while the considerable value of the NPCR and UACI criteria denotes the robustness of the approach against differential attacks. This algorithm has demonstrated superiority on both grayscale and color images.
This work proposes a memristor-based quantized convolutional auto-encoder (MQCAE) and applies it in an image denoising application. The pulse width or amplitude is gradually tuned by incremental steps in current memri...
详细信息
This work proposes a memristor-based quantized convolutional auto-encoder (MQCAE) and applies it in an image denoising application. The pulse width or amplitude is gradually tuned by incremental steps in current memristive programming methods, which can be extremely time-consuming. In this work, a multi-level programming method without incremental steps is proposed. MQCAE is composed of five kinds of network layers: convolution layers, deconvolution layers, activation function layers, batch normalization layers and max-pooling layers. We design a memristive circuit for realizing convolution and deconvolution with different kernel parameters. The proposed circuit generates one output feature row every cycle. Analog data buffers are designed to store the intermediate data among network layers. In addition, a reconfigurable analog circuit for realizing activation functions and batch normalization is presented. By using analog temp modules, convolution layers and max-pooling layers are able to compute simultaneously. We construct the MQCAE and introduce the pipeline technique among network layers based on the circuit modules. As a result, MQCAE is able to process one FashionMNIST image in 36 cycles with clock frequency 20 kHz. Finally, we verify the effectiveness of MQCAE in an image denoising task. The results show that the denoising performance of proposed scheme is close to software model while the processing speed is faster.
Lamb wave-based detection has become a promising technique for structural health monitoring in plate-like structures. However, the dispersion effect of Lamb waves makes the wave packets elongated, which degrades the r...
详细信息
Lamb wave-based detection has become a promising technique for structural health monitoring in plate-like structures. However, the dispersion effect of Lamb waves makes the wave packets elongated, which degrades the resolution for damage identification. Hence, it is necessary to develop effective methods for the dispersion removal of Lamb waves. Recently, deep learning has drawn much attention for Lamb wave detection due to its powerful feature extraction capability. Most current deep learning models are developed directly for damage classification or location estimation tasks on the Lamb wave signals, rather than signal processing for further manipulation. Therefore, in this paper, a novel approach based on a convolutional auto-encoder is proposed for the dispersion compensation of Lamb waves. The convolutional auto-encoder model is utilized to construct the mapping relationship between the dispersive signal and the time of flight of the wave packet. The dispersion compensated signal is reconstructed by combining the estimated time of flight from the network with the waveform of the excitation signal. Numerical and experimental validations on aluminum and composite plates are implemented to verify the effectiveness of the proposed method. Compared with conventional methods for dispersion compensation, the results demonstrate that our proposed method not only separates the overlapped wave packets but also enables dispersion compensation of multimodal and multi-packet Lamb wave signals. In addition, this method is still applicable in the case of inaccurate dispersion data due to the generalizability of the data-driven model.
In avionics , industrial electronic systems, analog circuits are one of the most commonly used components. Intermittent faults (IFs) are a no fault found (NFF) state in analog circuits that are difficult to detect. In...
详细信息
In avionics , industrial electronic systems, analog circuits are one of the most commonly used components. Intermittent faults (IFs) are a no fault found (NFF) state in analog circuits that are difficult to detect. In addition, the presence of noise may obscure critical information about the state of the circuit. Considering these challenges, this paper proposes an adaptive multiscale and dual subnet convolutional auto-encoder (AMDSCAE) to detect IFs. The proposed method can adaptively assign different attention to each scale and then fuse the multiscale information, which has better noise robustness. Then, the fault reconstruction error is amplified by the dual subnet structure to enhance the IF detection ability and find weaker faults. Considering the difficulty of obtaining fault sample labels, the proposed model requires only fault-free samples in the training process. In three typical analog filter circuit experiments, AMDSCAE has better noise immunity and can detect weaker IFs.(c) 2022 ISA. Published by Elsevier Ltd. All rights reserved.
The estimation of an image geo-site solely based on its contents is a promising task. Compelling image labelling relies heavily on contextual information, which is not as simple as recognizing a single object in an im...
详细信息
The estimation of an image geo-site solely based on its contents is a promising task. Compelling image labelling relies heavily on contextual information, which is not as simple as recognizing a single object in an image. An auto-Encode-based support vector machine approach is proposed in this work to estimate the image geo-site to address the issue of misclassifying the estimations. The proposed method for geo-site estimation is conducted using a dataset consisting of 125 classes of various images captured within 125 countries. The proposed work uses a convolutionalauto-Encode for training and dimensionality reduction. After that, the acquired preprocessed input dataset is further processed by a multi-label support vector machine. The performance assessment of the proposed approach has been accomplished using accuracy, sensitivity, specificity, and F1-score as evaluation parameters. Eventually, the proposed approach for image geo-site estimation presented in this article outperforms auto-Encode-based K-Nearest Neighbor and auto-Encode-Random Forest methods.
Recently, the attention mechanism has been effectively implemented in convolutional neural networks to boost performance of several computer vision tasks. Recognizing the potential of the attention mechanism in medica...
详细信息
Recently, the attention mechanism has been effectively implemented in convolutional neural networks to boost performance of several computer vision tasks. Recognizing the potential of the attention mechanism in medical imaging, we present an end-to-end-trainable spatial Attention based convolutional neural network architecture for recognizing diabetic retinopathy severity level. Initially spatial representations of the fundus images are projected to reduced space using a stacked convolutional auto-encoder. In order to enhance discrimination in reduced space, the auto-encoder is jointly trained with the classifier in an end-to-end manner. Attention mechanism introduced in the classification module ensures high emphasis on lesion regions compared to the non-lesion regions. The proposed model is evaluated on two benchmark datasets, and the experimental outcomes indicate that joint training favors stability and complements the learned representations when used along with attention. The proposed approach outperforms several existing models by achieving an accuracy of 84.17%, 63.24% respectively on Kaggle APTOS19 and IDRiD datasets. In addition, ablation studies validate our contributions and the behavior of the proposed model on both the datasets.
暂无评论