Single-image dehazing is a challenging task in computer vision research. Aiming at the limitations of traditional convolutional neural network representation capabilities and the high computational overhead of the self-attention mechanism in recent years, we proposed image attention and designed a single image dehazing network based on the image attention: IAD-Net. The proposed image attention is a plug-and-play module with the ability of global modeling. IAD-Net is a parallel network structure that combines the global modeling ability of image attention and the local modeling ability of convolution, so that the network can learn global and local features. The proposed network model has excellent feature learning ability and feature expression ability, has low computational overhead, and also improves the detail information of hazy images. Experiments verify the effectiveness of the image attention module and the competitiveness of IAD-Net with state-of-the-art methods.

To accurately extract tabular data, we propose a novel cell-based tabular data extraction model (TDEM). The key of TDEM is to utilize grayscale projection of row separation lines, coupled with table masks and column masks generated by the VGG-19 neural network, to segment each individual cell from the input image of the table. In this way, the text content of the table is extracted from a specific single cell, which greatly improves the accuracy of table recognition.

Based on the framework of a multi-stage key recovery attack for a large block cipher, 2 and 3-round differential-neural distinguishers were trained for AES using partial ciphertext bits. The study introduces the differential characteristics employed for the 2-round ciphertext pairs and explores the reasons behind the near 100% accuracy of the 2-round differential neural distinguisher. Utilizing the trained 2-round distinguisher, the 3-round subkey of AES is successfully recovered through a multi-stage key guessing. Additionally, a complexity analysis of the attack is provided, validating the effectiveness of the proposed method.

Chinese-Vietnamese cross-lingual event retrieval aims to retrieve the Vietnamese sentence describing the same event as a given Chinese query sentence from a set of Vietnamese sentences. Existing mainstream cross-lingual event retrieval methods rely on extracting textual representations from query texts and calculating their similarity with textual representations in other language candidate sets. However, these methods ignore the difference in event elements present during Chinese-Vietnamese cross-language retrieval. Consequently, sentences with similar meanings but different event elements may be incorrectly considered to describe the same event. To address this problem, we propose a cross-lingual retrieval method that integrates event elements. We introduce event elements as an additional supervisory signal, where we calculate the semantic similarity of event elements in two sentences using an attention mechanism to determine the attention score of the event elements. This allows us to establish a one-to-one correspondence between event elements in the text. Additionally, we leverage the multilingual pre-trained language model fine-tuned based on contrastive learning to obtain cross-language sentence representation to calculate the semantic similarity of the sentence texts. By combining these two approaches, we obtain the final text similarity score. Experimental results demonstrate that our proposed method achieves higher retrieval accuracy than the baseline model.

Chinese spelling correction is a foundational task in natural language processing that aims to detect and correct spelling errors in text. Most spelling corrections in Chinese used multimodal information to model the relationship between incorrect and correct characters. However, feature information mismatch occured during fusion result from the different sources of features, causing the importance relationships between different modalities to be ignored, which in turn restricted the model from learning in an efficient manner. To this end, this paper proposes a multimodal language model-based Chinese spelling corrector, named as MISpeller. The method, based on ChineseBERT as the basic model, allows the comprehensive capture and fusion of character semantic information, phonetic information and graphic information in a single model without the need to construct additional neural networks, and realises the phenomenon of unequal fusion of multi-feature information. In addition, in order to solve the overcorrection issues, the replication mechanism is further introduced, and the replication factor is used as the dynamic weight to efficiently fuse the multimodal information. The model is able to control the proportion of original characters and predicted characters according to different input texts, and it can learn more specifically where errors occur. Experiments conducted on the SIGHAN benchmark show that the proposed model achieves the state-of-the-art performance of the F1 score at the correction level by an average of 4.36%, which validates the effectiveness of the model.

Recent years have seen remarkable progress in human pose estimation. However, manual annotation of keypoints remains tedious and imprecise. To alleviate this problem, this paper proposes a novel method called Multi-Scale Contrastive Learning (MSCL). This method uses a siamese network structure with upper and lower branches that capture diffirent views of the same image. Each branch uses a backbone network to extract image representations, employing multi-scale feature vectors to capture information. These feature vectors are then passed through an enhanced feature pyramid for fusion, producing more robust feature representations. The feature vectors are then further encoded by mapping and prediction heads to predict the feature vector of another view. Using negative cosine similarity between vectors as a loss function, the backbone network is pre-trained on a large-scale unlabeled dataset, enhancing its capacity to extract visual representations. Finally, transfer learning is performed on a small amount of labelled data for the pose estimation task. Experiments on COCO datasets show significant improvements in Average Precision (AP) of 1.8%, 0.9%, and 1.2% with 1%, 5%, and 10% labelled data on COCO. In addition, the Percentage of Correct Keypoints (PCK) improves by 0.5% on MPII&AIC, outperforming mainstream contrastive learning methods.

End-to-end speech translation (ST) directly renders source language speech to the target language without intermediate automatic speech recognition (ASR) output as in a cascade approach. End-to-end ST avoids error propagation from intermediate ASR results. Although recent attempts have applied multi-task learning using an auxiliary task of ASR to improve ST performance, they use cross-entropy loss to one-hot references in the ASR task, and the trained ST models do not consider possible ASR confusion. In this study, we propose a novel multi-task learning framework for end-to-end STs leveraged by ASR-based loss against posterior distributions obtained using a pre-trained ASR model called ASR posterior-based loss (ASR-PBL). The ASR-PBL method, which enables a ST model to reflect possible ASR confusion among competing hypotheses with similar pronunciations, can be applied to one of the strong multi-task ST baseline models with Hybrid CTC/Attention ASR task loss. In our experiments on the Fisher Spanish-to-English corpus, the proposed method demonstrated better BLEU results than the baseline that used standard CE loss.

Due to the revitalization of the semiconductor industry and efforts to reduce labor and unmanned operations in the retail and food manufacturing industries, objects to be recognized at production sites are increasingly diversified in color and design. Depending on the target objects, it may be more reliable to process only color information, while intensity information may be better, or a combination of color and intensity information may be better. However, there are not many conventional method for optimizing the color and intensity information to be used, and deep learning is too costly for production sites. In this paper, we optimize the combination of the color and intensity information of a small number of pixels used for matching in the framework of template matching, on the basis of the mutual relationship between the target object and surrounding objects. We propose a fast and reliable matching method using these few pixels. Pixels with a low pixel pattern frequency are selected from color and grayscale images of the target object, and pixels that are highly discriminative from surrounding objects are carefully selected from these pixels. The use of color and intensity information makes the method highly versatile for object design. The use of a small number of pixels that are not shared by the target and surrounding objects provides high robustness to the surrounding objects and enables fast matching. Experiments using real images have confirmed that when 14 pixels are used for matching, the processing time is 6.3 msec and the recognition success rate is 99.7%. The proposed method also showed better positional accuracy than the comparison method, and the optimized pixels had a higher recognition success rate than the non-optimized pixels.

Before introducing systems to an activity in a business or in daily life, the effects of these systems should first be carefully examined by analysts. Thus, methods for examining such effects are required at the early stage of requirements analysis. In this study, we propose and evaluate an analysis method using a modeling notation for this purpose, called goal dependency modeling and analysis (GDMA). In an activity, an actor, such as a person or a system, expects a goal to be achieved. The actor or another actor will achieve this goal. We focus herein on such a goal and the two different roles played by the actors. In GDMA, the dependencies in the roles of the two actors about a goal are mainly represented. GDMA enables analysts to observe the change of actors, their expectations, and abilities by using metrics. Each metric is defined on the basis of the GDMA meta-model. Therefore, GDMA enables them to decide whether the change is good or bad both quantitatively and qualitatively for the people. We evaluate GDMA by describing models of the actual system introduction written in the literatures and explain the effects caused by this introduction. In addition, CASE tools are crucial in efficiently and accurately performing GDMA. Hence, we develop its tools by extending an existing UML modeling tool.

Reliability is an important figure of merit of the system and it must be satisfied in safety-critical applications. This paper considers parallel applications on heterogeneous embedded systems and proposes a two-phase algorithm framework to minimize energy consumption for satisfying applications’ reliability requirement. The first phase is for initial assignment and the second phase is for either satisfying the reliability requirement or improving energy efficiency. Specifically, when the application’s reliability requirement cannot be achieved via the initial assignment, an algorithm for enhancing the reliability of tasks is designed to satisfy the application’s reliability requirement. Considering that the reliability of initial assignment may exceed the application’s reliability requirement, an algorithm for reducing the execution frequency of tasks is designed to improve energy efficiency. The proposed algorithms are compared with existing algorithms by using real parallel applications. Experimental results demonstrate that the proposed algorithms consume less energy while satisfying the application’s reliability requirements.

In this work, we propose a 1T1R ReRAM CiM architecture for Hyperdimensional Computing (HDC). The number of Source Lines and Bit Lines is reduced by introducing memory cells that are connected in series, which is especially advantageous when using a 3D implementation. The results of CiM operations contain errors, but HDC is robust against them, so that even if the XNOR operation has an error of 25%, the inference accuracy remains above 90%.

Integrating RGB and event sensors improves object detection accuracy, especially during the night, due to the high-dynamic range of event camera. However, introducing an event sensor leads to an increase in computational resources, which makes the implementation of RGB-event fusion multi-modal AI to CiM difficult. To tackle this issue, this paper proposes RGB-Event fusion Multi-modal analog Computation-in-Memory (CiM), called REM-CiM, for multi-modal edge object detection AI. In REM-CiM, two proposals about multi-modal AI algorithms and circuit implementation are co-designed. First, Memory capacity-Efficient Attentional Feature Pyramid Network (MEA-FPN), the model architecture for RGB-event fusion analog CiM, is proposed for parameter-efficient RGB-event fusion. Convolution-less bi-directional calibration (C-BDC) in MEA-FPN extracts important features of each modality with attention modules, while reducing the number of weight parameters by removing large convolutional operations from conventional BDC. Proposed MEA-FPN w/ C-BDC achieves a 76% reduction of parameters while maintaining mean Average Precision (mAP) degradation to < 2.3% during both day and night, compared with Attentional FPN fusion (A-FPN), a conventional BDC-adopted FPN fusion. Second, the low-bit quantization with clipping (LQC) is proposed to reduce area/energy. Proposed REM-CiM with MEA-FPN and LQC achieves almost the same memory cells, 21% less ADC area, 24% less ADC energy and 0.17% higher mAP than conventional FPN fusion CiM without LQC.

In this work, fluctuation patterns of ReRAM current are classified automatically by proposed fluctuation pattern classifier (FPC). FPC is trained with artificially created dataset to overcome the difficulties of measured current signals, including the annotation cost and imbalanced data amount. Using FPC, fluctuation occurrence under different write conditions is analyzed for both HRS and LRS current. Based on the measurement and classification results, physical models of fluctuations are established.

Extremely low-voltage charge pump (ELV-CP) and its dedicated multi-stage driver (MS-DRV) for sub-60-mV thermoelectric energy harvesting are proposed. The proposed MS-DRV utilizes the output voltages of each ELV-CP to efficiently boost the control clock signals. The boosted clock signals are used as switching signals for each ELV-CP and MS-DRV to turn switch transistors on and off. Moreover, reset transistors are added to the MS-DRV to ensure an adequate non-overlapping period between switching signals. Measurement results demonstrated that the proposed MS-DRV can generate boosted clock signals of 350 mV from input voltage of 60 mV. The ELV-CP can boost the input voltage of 100 mV with 10.7% peak efficiency. The proposed ELV-CP and MS-DRV can boost the low input voltage of 56 mV.

This paper describes a programmable differential bandgap reference (PD-BGR) for ultra-low-power IoT (Internet-of-Things) edge node devices. The PD-BGR consists of a current generator (CG) and differential voltage generator (DVG). The CG is based on a bandgap reference (BGR) and generates an operating current and a voltage, while the DVG generates another voltage from the current. A differential voltage reference can be obtained by taking the voltage difference from the voltages. The PD-BGR can produce a programmable differential output voltage by changing the multipliers of MOSFETs in a differential pair and resistance with digital codes. Simulation results showed that the proposed PD-BGR can generate 25- to 200-mV reference voltages with a 25-mV step within a ±0.7% temperature inaccuracy in a temperature range from -20 to 100°C. A Monte Carlo simulation showed that the coefficient of the variation in the reference was within 1.1%. Measurement results demonstrated that our prototype chips can generate stable programmable differential output voltages, almost the same results as those of the simulation. The average power consumption was only 88.4 nW, with a voltage error of -4/+3 mV with 5 samples.

Recent years have seen a general resurgence of interest in analog signal processing and computing architectures. In addition, extensive theoretical and experimental literature on chaos and analog chaotic oscillators exists. One peculiarity of these circuits is the ability to generate, despite their structural simplicity, complex spatiotemporal patterns when several of them are brought towards synchronization via coupling mechanisms. While by no means a systematic survey, this paper provides a personal perspective on this area. After briefly covering design aspects and the synchronization phenomena that can arise, a selection of results exemplifying potential applications is presented, including in robot control, distributed sensing, reservoir computing, and data augmentation. Despite their interesting properties, the industrial applications of these circuits remain largely to be realized, seemingly due to a variety of technical and organizational factors including a paucity of design and optimization techniques. Some reflections are given regarding this situation, the potential relevance to discontinuous innovation in analog circuit design of chaotic oscillators taken both individually and as synchronized networks, and the factors holding back the transition to higher levels of technology readiness.

In this paper, we delve into wireless communications in the 300 GHz band, focusing in particular on the continuous bandwidth of 44 GHz from 252 GHz to 296 GHz, positioning it as a pivotal element in the trajectory toward 6G communications. While terahertz communications have traditionally been praised for the high speeds they can achieve using their wide bandwidth, focusing the beam has also shown the potential to achieve high energy efficiency and support numerous simultaneous connectivity. To this end, new performance metrics, EIRPλ and EINFλ, are introduced as important benchmarks for transmitter and receiver performance, and their consistency is discussed. We then show that, assuming conventional bandwidth and communication capacity, the communication distance is independent of carrier frequency. Located between radio waves and light in the electromagnetic spectrum, terahertz waves promise to usher in a new era of wireless communications characterized not only by high-speed communication, but also by convenience and efficiency. Improvements in antenna gain, beam focusing, and precise beam steering are essential to its realization. As these technologies advance, the paradigm of wireless communications is expected to be transformed. The synergistic effects of antenna gain enhancement, beam focusing, and steering will not only push high-speed communications to unprecedented levels, but also lay the foundation for a wireless communications landscape defined by unparalleled convenience and efficiency. This paper will discuss a future in which terahertz communications will reshape the contours of wireless communications as the realization of such technological breakthroughs draws near.

Considering the non-convexity of hybrid precoding and the hardware constraints of practical systems, a hybrid precoding architecture, which combines limited-resolution overlapped phase shifter networks with lens array, is investigated. The analogy part is a beam selection network composed of overlapped low-resolution phase shifter networks. In particular, in the proposed hybrid precoding algorithm, the analog precoding improves array gain by utilizing the quantization beam alignment method, whereas the digital precoding schemes multiplexing gain by adopting a Wiener Filter precoding scheme with a minimum mean square error criterion. Finally, in the sparse scattering millimeter-wave channel for the uniform linear array, the proposed method is compared with the existing scheme by computer simulation by using the ideal channel state information and the non-ideal channel state information. It is concluded that the proposed scheme performs better in low signal-to-noise regions and can achieve a good compromise between system performance and hardware complexity.

We propose microwave heating via electromagnetic coupling using zeroth-order resonators (ZORs) to extend the uniform heating area. ZORs can generate resonant modes with a wavenumber of 0, which corresponds to an infinite guide wavelength. Under this condition, uniform heating is expected because the resulting standing waves would not have nodes or antinodes. In the design proposed in this paper, two ZORs fabricated on dielectric substrates are arranged to face each other for electromagnetic coupling, and a sample placed between the resonators is heated. A single ZOR was investigated using a 3D electromagnetic simulator, and the resonant frequency and electric field distribution of the simulated ZOR were confirmed to be in good agreement with those of the fabricated ZOR. Simulations of two ZORs facing each other were then conducted to evaluate the performance of the proposed system as a heating apparatus. It was found that a resonator spacing of 25 mm was suitable for uniform heating. Heating simulations of SiC and Al2O3 sheets were performed with the obtained structure. The heating uniformity was evaluated by the width L50% over which the power loss distribution exceeds half the maximum value. This evaluation index was equal to 0.397λ0 for SiC and 0.409λ0 for Al2O3, both of which exceed λ0/4, the distance between a neighboring node and antinode of a standing wave, where λ0 is the free-space wavelength. Therefore, the proposed heating apparatus is effective for uniform microwave heating. Because of the different electrical parameters of the heated materials, SiC can be easily heated, whereas Al2O3 heats little. Finally, heating experiments were performed on each of these materials. Good uniformity in temperature was obtained for both SiC and Al2O3 sheets.

This paper proposes a novel configuration for a wideband single-carrier transmitter using a sub-terahertz frequency. For wideband single-carrier transmission over a bandwidth of several gigahertz, the frequency response non-flatness derived from transmitter components in an operating band seriously deteriorates the transmission quality due to inter-symbol interference. A promising approach to address this problem is equalizing the frequency response non-flatness at the transmitter. The proposed novel configuration has a feedback route for calculating the inverse frequency response and multiplying it with a transmission signal spectrum in the frequency domain. Moreover, we verify that employing the proposed transmitter configuration simplifies the receiver configuration by lowering the calculation complexity to minimize the inter-symbol interference to meet the signal-to-interference-and-noise ratio requirements. To confirm the feasibility of the proposed configuration, the transmission quality obtained using the proposed configuration is measured and evaluated. Experimental results confirm that the proposed configuration improves the error vector magnitude value to over 5 dB for a 10 Gbaud transmission and the transmission data rate of 25 Gbps.