Search

With the rapid development of miniaturization and miniaturization of portable electronic devices, the requirements of electronic devices are increasing for the performance of energy storage components within a certain volume. Considering the low power density and short cycle life of microbatteries, they cannot meet the requirements of rapid charge/discharge and long life of microelectronic devices. Meanwhile, microsupercapacitors (MSCs) have many advantages, such as high power density and long service life, and they meet the needs of use and development and have good prospects and progress advantages. In recent years, as new energy storage devices, MSCs have been widely studied in the design of device structure, but they are limited by many challenges such as low energy storage density, large internal resistance, improved cycle stability, and complex micropreparation technology. In this paper, the principle of MSCs and the research progress of structured electrodes were reviewed, thus providing a reference for the continuous optimization of the structure.

The implementation of IoT related technologies in tourism is promoting the achievement of smart tourism. In this context, Fifth generation (5G) technology, which aims to address the limitations of previous cellular systems, massively expands the implementation of IoT and gradually drives the Internet future to the edge. In order to facilitate the implementation, the work tries to formulate a definition of IoT enabled tourism and make a visualization review on the relevant literature to IoT enabled tourism in recent 10 years. The definition of IoT enabled tourism is helpful for extending the current research topics related to smart tourism. The visualization review identifies the intellectual bases and influential studies of IoT enabled tourism. Meanwhile, the outstanding cooperation relationships among authors, institutions and countries are detected out. These findings are helpful for the academic circle making further efforts on IoT enabled tourism and finally facilitating the achievement of smart tourism.

Semantic role labeling (SRL) aims at identifying the predicate-argument structure of a sentence. Recent work has significantly improved SRL performance by incorporating syntactic information and exploiting pre-trained models like BERT. Most of them use pre-trained models as isolated encoders to obtain word embeddings and enhance them with word-level syntax. Unlike many other languages, Chinese pre-trained models normally use Chinese characters instead of subwords as the basic input units, making the many-units-in-one-word phenomena more frequent and the relationship between characters more important. However, this character-level information is often ignored by previous research. In this paper, we propose the Character-Level Syntax-Infused network for Chinese SRL, which effectively incorporates the syntactic information between Chinese characters into pre-trained models. Experiments on the Chinese benchmarks of CoNLL-2009 and Universal Proposition Bank (UPB) show that the proposed approach achieves state-of-the-art results.

High-speed rail (HSR) has become a competitive mode with aviation for medium-distance intercity travel, given the massive deployment of the HSR infrastructure network in China. While the travel experience with both HSR and air has become more convenient, the systems’ operational reliability in terms of punctuality remains a key concern, especially during disruptive events, such as under severe weather conditions. Although previous studies have attempted to investigate the impact of severe weather events on the operational performance of transportation systems, there is still a lack of ability to forecast to what extent the performance of different transportation systems may vary under various conditions. This study develops an integrated modeling framework that allows us to predict the performance of weather-induced delays of different transportation systems, including HSR and aviation. By applying machine-learning methods to real-world transportation performance data, the study examines the robustness of the method, variations of data characteristics and the different applications of the predictive modeling system. Overall, the concept and modeling framework provide important implications for the improvement of transportation system resilience to various severe weather-related disruptions through the understanding of the impact and its predictability of the system performance.

Due to the limited training data, current data-driven algorithms, including deep convolutional networks (DCNs), are susceptible to training data that cannot be applied to new data directly. Unlike existing methods that are trying to improve model generation capability using limited data, we introduce a learning-based image translation method to generate data that share the same characteristics of target data. The low-resolution panchromatic satellite images are converted into high-resolution color images through interpolation and colorization with the proposed symmetric colorization network (SCN). Experiments on a very-high-resolution (VHR) dataset show that images generated by our SCN are with both quantitatively and qualitatively high color fidelity. Furthermore, we also demonstrate that high extraction accuracy is retained during the model transferring from aerial to satellite images. For pre-trained feature pyramid network (FPN), compared to the performance on raw panchromatic images, the interpolated and colorized images increase 305.7% of recall (0.929 vs. 0.229), 78.2% of overall accuracy (0.768 vs. 0.431), 132.5% of f1-score (0.851 vs. 0.366), and 230.8% of Jaccard index (0.741 vs. 0.224), respectively.

Bridge detection in aerial images is to determine whether a given aerial image contains one or more bridges and locate them. However, the arbitrary orientations, extreme aspect ratios, and variable backgrounds pose great challenges for bridge detection and positioning. In this article, we tackle these problems by combining the strengths of semantic-segmentation-based auxiliary supervision, waterbody constraint, and instance-switching-based data augmentation. More precisely, we make three main contributions. First, we propose an oriented bridge detection model with an auxiliary task of waterbody segmentation, which performs as guidance for bridge localization. The network is specifically designed in cascade style to handle the bridge detection and waterbody segmentation task end-to-end. Second, we make use of the semantic features of waterbody as spatial attention to distinguish bridges from cluttered backgrounds and then generate the waterbody segmentation map as the waterbody constraint, which introduces the prior knowledge of bridge distribution to refine the network predictions. Third, we propose a background consistent instance switching method for online data augmentation to further improve the robustness of bridge detection. To verify the effectiveness of the proposed method, we introduce a dataset named BridgeDetV1 containing 5000 well-annotated images with two kinds of bridge representations, i.e., the horizontal bounding box and the oriented bounding box. Extensive experiments demonstrate that our approach outperforms the state-of-the-art methods on this challenging benchmark. Dataset and code are available at https://github.com/whughw/BridgeDet.

Ride-hailing, as a popular shared-transportation method, has been operated in many areas all over the world. Researchers conducted various researches based on global cases. They argued on whether car-hailing is an effective travel mode for emission reduction and drew different conclusions. The detailed emission performance of the ride-hailing system depends on the cases. Therefore, there is an urgent demand to reduce the overall picking up distance during the dispatch. In this study, we try to satisfy this demand by proposing an optimization method combined with a prediction model to minimize the global void cruising distance when solving the dispatch problem. We use Didi ride-hailing data on one day for simulation and found that our method can reduce the picking up distance by 7.51% compared with the baseline greedy algorithm. The proposed algorithm additionally makes the average waiting time of passengers more than 4 minutes shorter. The statistical results also show that the performance of our method is stable. Almost the metric in all cases can be kept in a low interval. What is more, we did a day-to-day comparison. We found that, despite the different spatial-temporal distribution of orders and drivers on different day conditions, there are little differences in the performance of the method. We also provide temporal analysis on the changing pattern of void cruising distance and quantity of orders on weekdays and weekends. Our findings show that our method can averagely reduce more void cruising distance when ride-hailing is active compared with the traditional greedy algorithm. The result also shows that the method can stably reduce void cruising distance by about 4000 to 5000 m per order across one day. We believe that our findings can improve deeper insight into the mechanism of the ride-hailing system and contribute to further studies.

Internet of Things devices are popular in civilian and military applications, including smart device cities, smart grids, smart pipelines, and medical Internet of Things. Among them, carsharing supported by the Internet of Things is developing rapidly due to their advantages in environmental protection and reducing traffic congestion. The optimization of the carsharing system needs to consider the uncertainty of demand and the coupling relationship of multiple decision variables, which brings difficulties to the establishment of mathematical models and the design of efficient algorithms. Existing studies about carsharing optimization are mainly divided into four subproblems: the operation mode selection, vehicle type selection, demand analysis, or decision-making, rather than comprehensive consideration. This paper summarizes the four subproblems from the perspective of mathematical models, solving algorithms, and statistical methods and provides references for more comprehensive research in the future.

Low-rank representation (LRR) is a very competitive technique in various real-world applications for its powerful capability in discovering latent structure of noisy or corrupted data set. However, traditional low-rank models treat each data point and feature equally so that noisy data cannot be detected and suppressed effectively and have obvious deterioration in performance, especially in heavy noisy scenario. In this paper, to address this problem, we develop a method of feature and sample scaling for low rank representation. The importance of data points and their features in both feature and sample spaces are considered, as such, clean data points and noisy data points and their features can be distinguished. In addition, based on the observation that noisy data points are usually deviated far away from the principal projection of the data set, a cosine similarity metric between data vector and the principal projection vector is developed to measure the importance of each sample. Applying our method into two classical low rank models such as Low Rank Representation (LRR) and Bilinear Factorization (BF), we can learn better low-rank structure of clean data, while the outliers or missing data being suppressed. Extensive experimental results on ORL, COIL20 and video surveillance, demonstrate that our proposed method can outperform state-of-the-art low rank methods in image clustering tasks with various levels of corruptions, especially in a heavy noisy scenario.

Vehicle exhaust emissions at signalized intersections are the essential source of traffic-related pollution to pedestrians. Therefore, it is critical to predicting traffic emissions, especially the hazardous CO gas, with practical and accurate methods. However, the CO emission and concentration at crosswalks can be influenced by the complex traffic conditions in a complicated way, making the prediction of CO concentration a challenging task for traditional statistical models. To this end, a hybrid machine learning framework is proposed in this study to investigate the concentration of CO emissions at pedestrian crosswalks. The proposed method firstly ranks key influencing factors with a random forest approach. Then a prediction model with Multi-Variate Long Short-Term Memory (LSTM) neural networks based on the selected factors is developed. Data is collected at the field intersection for model training and validation. The autoregressive integrated moving average (ARIMA), support vector machines (SVM), radial basis functions network (RBFN), nonlinear vector autoregressive (VAR) and gated recurrent unit ( GRU ) neural network are selected as the benchmark models to verify the performance of the proposed model. The Root Mean Square Errors (RMSE), Mean Absolute Error (MAE) and R square are calculated to evaluate the performance of models comprehensively. The results indicated that the proposed model overwhelms the benchmark models in terms of prediction accuracy.

Navigation-assisted surgery is being studied and used by more and more doctors because of its high success rate, and many techniques are needed, among which diffusion tensor imaging (DTI) plays an important role. The navigation assisted surgery system integrated with DTI will be more perfect and powerful. In order to prove this, we first successfully made DTI images of patients in 3D Slicer. After careful observation and study, we found that in navigation assisted surgery, doctors can use DTI to observe the location of brain tissue more clearly. This paper mainly studies the principle and data formula of DTI, and compares and summarizes how the nerve bundle image generated by DTI plays a role in navigation assisted surgery. The method proposed is very workable for doctors, and the effect of operation is also very easy to see.

This article focuses on CT images of human brain. Based on the characteristics of CT, 3D Slicer software was used to segment the brain CT images. First, the functions and features of 3D Slicer software are briefly introduced. Second, the principles of threshold segmentation and FCM algorithm are described. Using the Segment Editor module of 3D Slicer software to perform image segmentation, the threshold segmentation method and FCM algorithm are compared.

In the history of natural language processing (NLP) development, the representation of words has always been a significant research topic. In this survey, we provide a comprehensive typology of word representation models from a novel perspective that the development from static to dynamic embeddings can effectively address the polysemy problem, which has been a great challenge in this field. Then the survey covers the main evaluation metrics and applications of these word embeddings. And, we further discuss the development of word embeddings from static to dynamic in cross-lingual scenario. Finally, we point out some open issues and future works.

This paper proposes a novel Piecewise Parabolic Approximate Computation method for hardware function evaluation, which mainly incorporates an error-flattened segmenter and an implementation quantizer. Under a required software maximum absolute error (MAE), the segmenter adaptively selects a minimum number of parabolas to approximate the objective function. By completely imitating the circuit’s behavior before actual implementation, the quantizer calculates the minimum quantization bit width to ensure a non-redundant fixed-point hardware architecture with an MAE of 1 unit of least precision (ulp), eliminating the iterative design time for the circuits. The method causes the number of segments to reach the theoretical limit, and has great advantages in the number of segments and the size of the look-up table (LUT). To prove the superiority of the proposed method, six common functions were implemented by the proposed method under TSMC-90 nm technology. Compared to the state-of-the-art piecewise quadratic approximation methods, the proposed method has advantages in the area with roughly the same delay. Furthermore, a unified function-evaluation unit was also implemented under TSMC-90 nm technology.

The haze in remote sensing images can cause the decline of image quality and bring many obstacles to the applications of remote sensing images. Considering the non-uniform distribution of haze in remote sensing images, we propose a single remote sensing image dehazing method based on the encoder–decoder architecture, which combines both wavelet transform and deep learning technology. To address the clarity issue of remote sensing images with non-uniform haze, we preliminary process the input image by the dehazing method based on the atmospheric scattering model, and extract the first-order low-frequency sub-band information of its 2D stationary wavelet transform as an additional channel. Meanwhile, we establish a large-scale hazy remote sensing image dataset to train and test the proposed method. Extensive experiments show that the proposed method obtains greater advantages over typical traditional methods and deep learning methods qualitatively. For the quantitative aspects, we take the average of four typical deep learning methods with superior performance as a comparison object using 500 random test images, and the peak-signal-to-noise ratio (PSNR) value using the proposed method is improved by 3.5029 dB, and the structural similarity (SSIM) value is improved by 0.0295, respectively. Based on the above, the effectiveness of the proposed method for the problem of remote sensing non-uniform dehazing is verified comprehensively.

Convolutional Neural Networks (CNNs) are effective and mature in the field of classification, while Spiking Neural Networks (SNNs) are energy-saving for their sparsity of data flow and event-driven working mechanism. Previous work demonstrated that CNNs can be converted into equivalent Spiking Convolutional Neural Networks (SCNNs) without obvious accuracy loss, including different functional layers such as Convolutional (Conv), Fully Connected (FC), Avg-pooling, Max-pooling, and Batch-Normalization (BN) layers. To reduce inference-latency, existing researches mainly concentrated on the normalization of weights to increase the firing rate of neurons. There are also some approaches during training phase or altering the network architecture. However, little attention has been paid on the end of inference phase. From this new perspective, this paper presents 4 stopping criterions as low-cost plug-ins to reduce the inference-latency of SCNNs. The proposed methods are validated using MATLAB and PyTorch platforms with Spiking-AlexNet for CIFAR-10 dataset and Spiking-LeNet-5 for MNIST dataset. Simulation results reveal that, compared to the state-of-the-art methods, the proposed method can shorten the average inference-latency of Spiking-AlexNet from 892 to 267 time steps (almost 3.34 times faster) with the accuracy decline from 87.95 to 87.72%. With our methods, 4 types of Spiking-LeNet-5 only need 24–70 time steps per image with the accuracy decline not more than 0.1%, while models without our methods require 52–138 time steps, almost 1.92 to 3.21 times slower than us.

A new adaptive model predictive control (AMPC) is proposed in this paper to enhance the performance of proton exchange membrane fuel cell (PEMFC) in DC microgrids. Compared to the existing methods, such as the PI control and conventional model predictive control, the proposed algorithm produces better tracking performances and overcomes the problem of model dependence. The effectiveness of the proposed algorithm under system parameter variations are verified by simulation results.

Hybrid AC/DC microgrids are increasingly used in power systems for remote industrial, commercial and residential plants. To maintain low generation cost while reducing carbon emissions, the combinational usage of conventional gas turbine generator and solar photovoltaic (PV) is usually preferred, where the problems of unsatisfying fuel efficiency and renewable intermittency are identified as the key challenges. To enhance the fuel efficiency of microgrid and effectively mitigate renewable intermittency, this paper proposes an advanced energy management strategy. It ensures high fuel efficiency and produces a virtual inertia for the microgrid. Moreover, fast DC-link voltage regulation and seamless transition between grid-forming and grid-following modes can be achieved. The stability of controllers employed in the proposed strategy is proven. The efficacy of the proposed energy management strategy is verified by simulation results.

This paper proposes a very cost-effective digital twin real-time hybrid simulation platform for engineering education in renewable energy areas. Compared to many of the existing laboratory setups that are designed for renewable engineering education, the proposed platform provides greater flexibility, better skill training opportunities, and excellent system extendibility. Its power hardware-in-the-loop (PHIL) based hybrid simulation results are desirable for industry applications. Furthermore, the proposed platform can receive real-time information of renewable energy systems through ethernet based communication. The superior performance and flexibility of the proposed platform are verified by experiments.

The current service composition problem is usually aimed at the constrained service composition problem of Web services. However, this kind of scheme is not suitable for the situation that one service corresponds to multiple tasks. In order to solve the problem of service composition meeting business constraints and users’ needs in a specific field, this paper uses Markov decision process to model the problem, proposes a method based on deep Q-learning to solve the problem, and uses random sampling method for the inference. This method calculates the candidate services that can be used for composition, and takes the maximum cumulative rewards represented by the degree of constraint satisfaction as the optimization objective, and globally optimizes the results to meet the needs of users to the greatest extent. The experimental results show that: in this problem, compared with the existing methods, this method has higher combination efficiency.

Integrated Energy Systems have become a vital energy utilization to alleviate the multiple stress of energy, environment, and economy worldwide. Integrated Energy Microgrid (IEM) is a small-scale integrated energy system located in a distribution network close to the demand side. The accurate forecasting of multi-load is an essential prerequisite for ensuring the reliable and economic operation of an IEM. Comprehensively considering temperature, humidity, wind speed, and the coupling relationship of multi-energy, this paper proposes a CNN-Seq2Seq model with an attention mechanism based on a multi-task learning method for a short-time multi-energy load forecasting. In detail, CNN is used to extract useful features of the input data. Then, the short-time multi-energy load is forecasted by using Seq2Seq according to the extracted features. Meanwhile, the attention mechanism and multi-task learning method are introduced to improve the accuracy of load forecasting. The simulation results with the actual data of an IEM validate the effectiveness of the proposed short-time multi-energy load forecasting method.

Subcutaneous injections are administered into the region beneath the skin, while avoiding puncturing the blood vessels, which gives many types of medications for various medical conditions. In this paper, a portable robotic system performing autonomous cannulation of subcutaneous injections is proposed, and it achieves to automatically locate the proper injection site by analyzing near-infrared (NIR) image sequences. The robot mainly consists of two functional modules-image processing module and motion control module. The former with a full-search algorithm processes the images obtained by the NIR equipment. The puncture point is selected in the area where there are no blood vessels and a method of “range square” is utilized. The motion control module of the robotic system employs the pulse width modulation (PWM) wave to effectively control the motors and manipulates the syringe to puncture at the selected point. The image processing algorithm was evaluated based on the real NIR images of volunteers’ hands and forearms, and the image servo control of the robot was tested on the phantom. The experimental results were analyzed by a medical professional, and the success rate of the image processing algorithm is 96.09%, while the puncture time can satisfy the clinical demand for the efficiency of puncture procedures.

Acne is an incredibly common skin condition caused by oil from clog hair follicles and dead skin. It is usually found in adolescence to young adulthood and may affects people of all ages. The diagnosis of acne usually requires manual examination by a well-trained dermatologist, which can take a lot of effort. Therefore, it is necessary to design an automatic classification algorithm for acne lesion. However, due to the lack of proper imaging method, acne-specific facial image analysis is still a difficult task. To address this issue we propose a novel approach using high-definite VISIA image. By incorporating better image and better models, an overall accuracy above 80% is achieved on a large-scale dataset consists of more than one thousand people. The results imply that automatic acne classification is a promising direction.

Mobile edge computing (MEC) is proposed as a new paradigm to meet the ever-increasing computation requirements, which is caused by the rapid growth of the Internet of Things (IoT) devices. As a supplement to the terrestrial network, satellites can provide communication to terrestrial devices in some harsh environments and natural disasters. Satellite edge computing is becoming an emerging topic and technology. In this paper, a game-theoretic approach to the optimization of computation offloading strategy in satellite edge computing is proposed. The system model for computation offloading in satellite edge computing is established, considering the intermittent terrestrial-satellite communication caused by satellites orbiting. We conduct a computation offloading game framework and compute the response time and energy consumption of a task based on the queuing theory as metrics of optimizing performance. The existence and uniqueness of the Nash equilibrium is theoretically proved, and an iterative algorithm is proposed to find the Nash equilibrium. Simulation results validate the proposed algorithm and show that the game-based offloading strategy can greatly reduce the average cost of a device.

Deep contextualized word embeddings (Embeddings from Language Model, short for ELMo), as an emerging and effective replacement for the static word embeddings, have achieved success on a bunch of syntactic and semantic NLP problems. However, little is known about what is responsible for the improvements. In this article, we focus on the effect of ELMo for a typical syntax problem—universal POS tagging and dependency parsing. We incorporate ELMo as additional word embeddings into the state-of-the-art POS tagger and dependency parser, and it leads to consistent performance improvements. Experimental results show the model using ELMo outperforms the state-of-the-art baseline by an average of 0.91 for POS tagging and 1.11 for dependency parsing. Further analysis reveals that the improvements mainly result from the ELMo’s better abstraction ability on the out-of-vocabulary (OOV) words, and the character-level word representation in ELMo contributes a lot to the abstraction. Based on ELMo’s advantage on OOV, experiments that simulate low-resource settings are conducted and the results show that deep contextualized word embeddings are effective for data-insufficient tasks where the OOV problem is severe.

The increasing population in the world forces humans to improve farm yields using advanced technologies. The Internet of Things (IoT) is one promising technique to achieve precision agriculture, which is expected to greatly increase yields. However, the large-scale application of IoT systems in agriculture is facing challenges such as huge investment in agriculture IoT systems and non-tech-savvy farmers. To identify these challenges, we summarize the applications of IoT techniques in agriculture in four categories: controlled environment planting, open-field planting, livestock breeding, and aquaculture and aquaponics. The focus on implementing agriculture IoT systems is suggested to be expanded from the growth cycle to the agri-products life cycle. Meanwhile, the energy concern should be considered in the implementation of agriculture IoT systems. The construction of green IoT systems in the whole life cycle of agri-products will have great impact on farmers' interest in IoT techniques. With the life cycle framework, emerging finance, operation, and management (FOM) issues in the implementation of green IoT systems in agriculture are observed, such as IoT finance, supply chain and big data financing, network nodes recharging and repairing, and IoT data management. These FOM issues call for innovative farm production modes and new types of agribusiness enterprises.

This paper presents an invisible and robust watermarking method and its hardware implementation. The proposed architecture is based on the discrete cosine transform (DCT) algorithm. Novel techniques are applied as well to reduce the computational cost of DCT and color space conversion to achieve low-cost and high-speed performance. Besides, a watermark embedder and a blind extractor are implemented in the same circuit using a resource-sharing method. Our approach is compatible with various watermarking embedding ratios, such as 1/16 and 1/64, with a PSNR of over 45 and the NC value of 1. After Joint Photographic Experts Group (JPEG) compression with a quality factor (QF) of 50, our method can achieve an NC value of 0.99. Results from a design compiler (DC) with TSMC-90 nm CMOS technology show that our design can achieve the frequency of 2.32 GHz with the area consumption of 304,980.08 μm2 and power consumption of 508.1835 mW. For the FPGA implementation, our method achieved a frequency of 421.94 MHz. Compared with the state-of-the-art works, our design improved the frequency by 4.26 times, saved 90.2% on area and increased the power efficiency by more than 1000 fold.

Traffic Forecasting is an important and challenging problem. The recent developed deep learning models are becoming dominant in this area. Especially, graph neural networks (GNNs) are being applied in traffic forecasting in recent years. In this paper, I give a review of the related work and the applications of GNNs in different traffic forecasting problems, e.g., bike sharing, metro flow, road traffic flow prediction, etc. I find that GNNs are only applied in recent years, and there is still a great research potential for this direction.

Fault diagnosis is an important way to ensure the operation security in complex industrial processes. Considering the inherent multiscale characteristics and time dependency about industrial process monitoring data, a novel fault diagnosis method based on multiscale temporal convolutional network (MS-TCN) was proposed in this paper. Firstly, different from the widely used time-domain features with one single scale, the multiscale time-frequency information extracted with the discrete wavelet transform was also introduced to represent the raw data. And a temporal convolutional network was then combined to capture longer-term temporal feature from the sequential processing data. The experimental results on the Tennessee Eastman process indicated that, our proposed method outperformed these state-of-the-art fault diagnosis methods, especially for the 3 incipient faults hard to classify.

Fault detection is an important part of process control in order to achieve the safe operation. Data-driven methods including Independent Component Analysis (ICA) have been widely studied and applied. ICA performs reliable analysis by maximizing non-Gaussianity and obtaining statistically independent latent variables. To improve the adaptability of ICA on the industrial data, a Log-normal Independent Component Analysis (LNICA) method was proposed in this paper to maximize non-Gaussianity of the observed data. The experimental results on digital simulation data showed that our method could effectively detect the pulse disturbance. And the results on Tennessee Eastman (TE) process indicated that the LNICA had less false alarm rate, better fault detection performance than the traditional PCA and ICA methods. Also the LNICA had higher fault detection efficiency in the independent component optimization.

Images taken in low-light environment often face with low dynamic range or color shift caused by lack of illumination. Although many algorithms have been proposed to enhance these images, not much effort has been made on quality assessment of these enhancement results. In our work, we built a database which contains 30 low-light images (both outdoor and indoor scene included) and enhanced images processed by 11 enhancement algorithms. We conducted a subjective experiment based on this database. We found that no algorithm can behave best with images in all situations. Generally, learning-based methods behave better. Further, we also did objective assessment on the database.

This paper presents ByteSing, a Chinese singing voice synthesis (SVS) system based on duration allocated Tacotron-like acoustic models and WaveRNN neural vocoders. Different from the conventional SVS models, the proposed ByteSing employs Tacotron-like encoder-decoder structures as the acoustic models, in which the CBHG models and recurrent neural networks (RNNs) are explored as encoders and decoders respectively. Meanwhile an auxiliary phoneme duration prediction model is utilized to expand the input sequence, which can enhance the model controllable capacity, model stability and tempo prediction accuracy. WaveRNN neural vocoders are also adopted as neural vocoders to further improve the voice quality of synthesized songs. Both objective and subjective experimental results prove that the SVS method proposed in this paper can produce quite natural, expressive and high-fidelity songs by improving the pitch and spectrogram prediction accuracy and the models using attention mechanism can achieve best performance., Comment: Accepted by ISCSLP2021

This paper presents a novel supervised approach to detecting the chorus segments in popular music. Traditional approaches to this task are mostly unsupervised, with pipelines designed to target some quality that is assumed to define "chorusness," which usually means seeking the loudest or most frequently repeated sections. We propose to use a convolutional neural network with a multi-task learning objective, which simultaneously fits two temporal activation curves: one indicating "chorusness" as a function of time, and the other the location of the boundaries. We also propose a post-processing method that jointly takes into account the chorus and boundary predictions to produce binary output. In experiments using three datasets, we compare our system to a set of public implementations of other segmentation and chorus-detection algorithms, and find our approach performs significantly better., Comment: This version is a preprint of an accepted paper by ICASSP2021. Please cite the publication in the Proceedings of IEEE International Conference on Acoustics, Speech, & Signal Processing

Speech enhancement is a task to improve the intelligibility and perceptual quality of degraded speech signal. Recently, neural networks based methods have been applied to speech enhancement. However, many neural network based methods require noisy and clean speech pairs for training. We propose a speech enhancement framework that can be trained with large-scale weakly labelled AudioSet dataset. Weakly labelled data only contain audio tags of audio clips, but not the onset or offset times of speech. We first apply pretrained audio neural networks (PANNs) to detect anchor segments that contain speech or sound events in audio clips. Then, we randomly mix two detected anchor segments containing speech and sound events as a mixture, and build a conditional source separation network using PANNs predictions as soft conditions for speech enhancement. In inference, we input a noisy speech signal with the one-hot encoding of "Speech" as a condition to the trained system to predict enhanced speech. Our system achieves a PESQ of 2.28 and an SSNR of 8.75 dB on the VoiceBank-DEMAND dataset, outperforming the previous SEGAN system of 2.16 and 7.73 dB respectively., Comment: 5 pages

Code generation is an important research field of software engineering, aiming to reduce development costs and improve program quality. Nowadays, more and more researchers intend to implement code generation by natural language understanding. In this paper, we propose a generation method to convert natural language descriptions to the program code based on deep learning. We use an encoder-decoder model with gated attention mechanism. Here, the decoder is an InterAction-LSTM. The gated attention combines the previous decoding cell state with source representations to improve the limitation of invariant source representations. The decoder makes the information interact each other before putting them into the gate of the LSTM. The code generation is verified on two datasets, Conala and Django. Compared with known models, our model outperforms the baselines both in Accuracy and Bleu.

Separating a song into vocal and accompaniment components is an active research topic, and recent years witnessed an increased performance from supervised training using deep learning techniques. We propose to apply the visual information corresponding to the singers' vocal activities to further improve the quality of the separated vocal signals. The video frontend model takes the input of mouth movement and fuses it into the feature embeddings of an audio-based separation framework. To facilitate the network to learn audiovisual correlation of singing activities, we add extra vocal signals irrelevant to the mouth movement to the audio mixture during training. We create two audiovisual singing performance datasets for training and evaluation, respectively, one curated from audition recordings on the Internet, and the other recorded in house. The proposed method outperforms audio-based methods in terms of separation quality on most test recordings. This advantage is especially pronounced when there are backing vocals in the accompaniment, which poses a great challenge for audio-only methods.

Traditional satellites are designed and developed according to specific functions, resulting in large size, high price and long development cycle. With the rapid development of small satellite technology, the satellite has higher and higher degree of modularization. Similar to smartphones, satellites can dynamically upload “Apps” in-orbit, achieving the transition from “function satellites” to “smart satellites”. In view of the rapid, efficient and reliable development of on-board software, a model-driven software development framework and a development tool chain are proposed in this paper. To solve the problems of lack of standardized architecture in on-board software development, poor communication of various development stages, serious coupling of software and hardware, and low automation, the framework adopts unified architecture, standardized components, configurable integration and automatic code generation. The development tool chain provides a complete set of tools for entire on-board software development based on the model-driven framework. It improves the software reusability by decoupling software design from hardware platform and shortens the development period by automatically connecting the various development stages. Finally, this paper demonstrates and assesses the process of developing iSat-1, which is a CubeSat for function in-orbit defined experiment.

Humans understand videos from both the visual and audio aspects of the data. In this work, we present a self-supervised cross-modal representation approach for learning audio-visual correspondence (AVC) for videos in the wild. After the learning stage, we explore retrieval in both cross-modal and intra-modal manner with the learned representations. We verify our experimental results on the VGGSound dataset [1], and our approach achieves promising results.

Support vector machine (SVM), which serves as one kind of artificial intelligence technique, has been widely employed in transformer fault diagnosis when involving dissolved gas analysis (DGA). However, when using SVM, it is easy to misclassify samples which are located near the decision boundary, resulting in a decrease in the accuracy of fault diagnosis. Given this issue, this paper proposed a genetic algorithm (GA) optimized probabilistic SVM (GAPSVM) integrated with the fuzzy three-ratio (FTR) method, in which the GAPSVM can judge whether a sample is near the decision boundary according to its output probabilities and diagnose the samples which are not near the decision boundary. Then, FTR is used to diagnose the samples which are near the decision boundary. Combining GAPSVM and FTR, the integrated model can accurately diagnose samples near the decision boundary of SVM. In addition, to avoid redundant and erroneous features, this paper also used GA to select the optimal DGA features. The diagnostic accuracy of the proposed GAPSVM integrated with the FTR fault diagnosis method reached 86.80% after 10 repeated calculations using 118 groups of IEC technical committee (TC) 10 samples. Moreover, the robustness is also proven through 30 groups of DGA samples from the State Grid Co. of China and 15 practical cases with missing values.

Integrated energy load forecasting is a prerequisite for integrated energy system (IES) planning and operation, therefore, accurate and rapid integrated energy load forecasting has important practical value. This paper first introduces the coupling and complementary relationship between different forms of energy in the integrated energy system, then the structure of LSTM network neural unit model is introduced, furthermore, a short-term multi-load forecasting method based on deep LSTM network is proposed, which method includes the construction of deep neural network model, the preprocessing of load and weather input, the evaluation index of root mean squared error(RMSE) and the selection of optimal global parameters based on random search method. Finally, actual data is applied to verify the effectiveness of the proposed method. After the comparison with other load forecasting method, The deep LSTM network multi-load prediction method proposed can obtain more accurate results and is suitable for practical engineering applications

Sinusoidal vibration is the most severe mechanical environment experienced during the mission period of the satellite laser communication terminal telescope, and it is also the difficulty and emphasis of the prediction of the environment of the satellite laser communication terminal telescope. Based on the satellite laser communication terminal telescope sine vibration test appraisal level conditions using Solidworks software model of the organization of the telescope, for X, Y, Z three directions of sine sweep, finite element analysis software can be simulated by harmonic response analysis, through the way of input frequency sweep acceleration amplitude, the sinusoidal response to a certain frequency range is analyzed. Sinusoidal analysis and impact response analysis show that the internal deformation of the model is small and the stress is less than the allowable stress of the structural materials. When subjected to impact response, the maximum stress of glass lens is 21Mpa. Meet the use requirements under the given mechanical environmental conditions.

For efficient building outline extraction, many algorithms, including unsupervised or supervised, have been proposed over the past decades. In recent years, due to the rapid development of the convolutional neural networks, especially fully convolutional networks, building extraction is treated as a semantic segmentation task that deals with the extremely biased positive pixels. The state-of-the-art methods, either through direct or indirect approaches, are mainly focused on better network design. The shifts and rotations, which are coarsely presented in manually created annotations, have long been ignored. Due to the limited number of positive samples, the misalignment will significantly reduce the correctness of pixel-to-pixel loss that might lead to a gradient explosion. To overcome this, we propose a nearest feature selector (NFS) to dynamically re-align the prediction and slightly misaligned annotations. The NFS can be seamlessly appended to existing loss functions and prevent misleading by the errors or misalignment of annotations. Experiments on a large scale aerial image dataset with centered buildings and corresponding building outlines indicate that the additional NFS brings higher performance when compared to existing naive loss functions. In the classic L1 loss, the addition of NFS gains increments of 8.8% of f1-score, 8.9% of kappa coefficient, and 9.8% of Jaccard index, respectively.

Edit-distance-based string similarity search has many applications such as spell correction, data de-duplication, and sequence alignment. However, computing edit distance is known to have high complexity, which makes string similarity search challenging for large datasets. In this paper, we propose a deep learning pipeline (called CNN-ED) that embeds edit distance into Euclidean distance for fast approximate similarity search. A convolutional neural network (CNN) is used to generate fixed-length vector embeddings for a dataset of strings and the loss function is a combination of the triplet loss and the approximation error. To justify our choice of using CNN instead of other structures (e.g., RNN) as the model, theoretical analysis is conducted to show that some basic operations in our CNN model preserve edit distance. Experimental results show that CNN-ED outperforms data-independent CGK embedding and RNN-based GRU embedding in terms of both accuracy and efficiency by a large margin. We also show that string similarity search can be significantly accelerated using CNN-based embeddings, sometimes by orders of magnitude., Comment: Accepted by the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, 2020

With the application of Internet of Things (IoT) in tourism, the functions and management modes of tourist attractions are being greatly updated. It becomes a faced problem to assess the intelligence level of IoT-based tourist attractions. The assessment is helpful for managers to equip their tourist attractions with smart services which further improve the management efficiency and tourist satisfaction. However, there are few recognized standards for the implementation of IoT-based tourist attractions, and the common practice of using the average value to replace multiple assessment scores has a shortage. Motivated by these observations, we present a framework of IoT-based intelligent tourist attractions and recognize specific intelligent functions brought by IoT techniques to tourist attractions. Then, two fuzzy TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) approaches, that is, a centroid-based fuzzy TOPSIS and an integral-based fuzzy TOPSIS, are formulated to deal with the inconsistent assessment scores from multiple experts. An application study shows the effectiveness and advantage of our approaches in comparison with the classical TOPSIS. Both the centroid-based fuzzy TOPSIS and the classical TOPSIS cannot reflect the preferences of decision-makers, but their assessment results are not fully consistent. The assessment results by the integral-based fuzzy TOPSIS are subject to the given optimism level which may make difference on assessment orders. We observe some insightful findings helpful for improving the intelligence level of IoT-based tourist attractions.

Source separation is the task to separate an audio recording into individual sound sources. Source separation is fundamental for computational auditory scene analysis. Previous work on source separation has focused on separating particular sound classes such as speech and music. Many of previous work require mixture and clean source pairs for training. In this work, we propose a source separation framework trained with weakly labelled data. Weakly labelled data only contains the tags of an audio clip, without the occurrence time of sound events. We first train a sound event detection system with AudioSet. The trained sound event detection system is used to detect segments that are mostly like to contain a target sound event. Then a regression is learnt from a mixture of two randomly selected segments to a target segment conditioned on the audio tagging prediction of the target segment. Our proposed system can separate 527 kinds of sound classes from AudioSet within a single system. A U-Net is adopted for the separation system and achieves an average SDR of 5.67 dB over 527 sound classes in AudioSet., 5 pages

This paper presents a COordinate Rotation Digital Computer (CORDIC)-based architecture for the computation of Nth root and proves its feasibility by hardware implementation. The proposed architecture performs the task of Nth root simply by shift-add operations and enables easy tradeoff between the speed (or precision) and the area. Technically, we divide the Nth root computation into three different subtasks, and map them onto three different classes of the CORDIC accordingly. To overcome the drawback of narrow convergence range of the CORDIC algorithm, we adopt several innovative methods to yield a much improved convergence range. Subsequently, in terms of convergence range and precision, a flexible architecture is developed. The architecture is validated using MATLAB with extensive vector matching. Finally, using a pipelined structure with fixed-point input data, we implement the example circuits of the proposed architecture with radicand ranging from zero to one million, and achieve an average mean of approximately 10−7 for the relative error. The design is modeled using Verilog HDL and synthesized under the TSMC 40-nm CMOS technology. The report shows a maximum frequency of 2.083 GHz with $197421.00~{\mu }\text{m}^{2}$ area. The area decreases to $169689.98~{\mu }\text{m}^{2}$ when the frequency lowers to 1.00 GHz.