ISSN 
Latest Articles
Articles in press have been peer-reviewed and accepted, which are not yet assigned to volumes/issues, but are citable by Digital Object Identifier (DOI).
, Available online , doi: 10.12178/1001-0548.2024156
Abstract:
In recent years, the rapid development of deep learning technology has greatly promoted the progress of virtual digital human technology, especially in the area of audio-driven digital human video generation. Research in this field has shown broad application prospects in various scenarios such as video translation, film production, and virtual assistants. The current methods and research status of speech-driven digital human video generation are sorted out and summarized, focusing on the key technologies, datasets, and evaluation strategies. In terms of key technologies, artificial intelligence technologies such as generative adversarial networks, diffusion models, and neural radiance fields have all played an important role. The scale and diversity of datasets are crucial for model training, and the improvement of evaluation strategies helps to evaluate the generation effect more objectively. The technology of virtual digital human video generation will continue to face numerous challenges and opportunities. It is expected that this field can continue to innovate and develop, bringing more convenience and fun to human society.
In recent years, the rapid development of deep learning technology has greatly promoted the progress of virtual digital human technology, especially in the area of audio-driven digital human video generation. Research in this field has shown broad application prospects in various scenarios such as video translation, film production, and virtual assistants. The current methods and research status of speech-driven digital human video generation are sorted out and summarized, focusing on the key technologies, datasets, and evaluation strategies. In terms of key technologies, artificial intelligence technologies such as generative adversarial networks, diffusion models, and neural radiance fields have all played an important role. The scale and diversity of datasets are crucial for model training, and the improvement of evaluation strategies helps to evaluate the generation effect more objectively. The technology of virtual digital human video generation will continue to face numerous challenges and opportunities. It is expected that this field can continue to innovate and develop, bringing more convenience and fun to human society.
, Available online , doi: 10.12178/1001-0548.2022360
Abstract:
Quantum computers have gradually entered people's vision and are expected to be able to efficiently solve problems that classical computers cannot solve. Quantum computers have great potential. This paper first summarizes the development of quantum computer systems (involving hardware, software and other levels), and summarizes the existing or need to face problems of current quantum computing systems. Then, the design principles of quantum operating system are presented, and the core mechanism of a prototype quantum operating system QuOS is introduced.
Quantum computers have gradually entered people's vision and are expected to be able to efficiently solve problems that classical computers cannot solve. Quantum computers have great potential. This paper first summarizes the development of quantum computer systems (involving hardware, software and other levels), and summarizes the existing or need to face problems of current quantum computing systems. Then, the design principles of quantum operating system are presented, and the core mechanism of a prototype quantum operating system QuOS is introduced.
, Available online , doi: 10.12178/1001-0548.2023273
Abstract:
Phishing, as a form of social engineering attack, aims to deceive victims by masquerading as a trustworthy entity such as a bank, social media platform, or government agency, using false emails, websites, or messages. Researchers primarily employ various technological means to detect phishing attacks, yet current detection studies face three main challenges. Firstly, attackers employ disguise, exploit vulnerabilities, and employ evasion techniques to evade detection. Secondly, existing detection methods suffer from poor interpretability, low real-time capabilities, and issues like concept drift. Lastly, due to insufficient interpretability, users may lack trust in the detection results. This summary comprehensively considers application scenarios, datasets, detection methods, etc., in current detection research, outlines the challenges faced, and anticipates potential future research directions.
Phishing, as a form of social engineering attack, aims to deceive victims by masquerading as a trustworthy entity such as a bank, social media platform, or government agency, using false emails, websites, or messages. Researchers primarily employ various technological means to detect phishing attacks, yet current detection studies face three main challenges. Firstly, attackers employ disguise, exploit vulnerabilities, and employ evasion techniques to evade detection. Secondly, existing detection methods suffer from poor interpretability, low real-time capabilities, and issues like concept drift. Lastly, due to insufficient interpretability, users may lack trust in the detection results. This summary comprehensively considers application scenarios, datasets, detection methods, etc., in current detection research, outlines the challenges faced, and anticipates potential future research directions.
Research progress and development trend of radio frequency/microwave power amplifier chip technology
, Available online , doi: 10.12178/1001-0548.2023266
Abstract:
Based on a comprehensive review and classification of the concepts, types, and realization processes of radio frequency/microwave power amplifier chips, the research status and urgent technical problems of key techniques such as high frequency, linearity improvement, energy conversion efficiency improvement, bandwidth expansion and highly integrated packaging and so on, are focused on. The mainstream realization of each key technology, typical cases of research and development, as well as the advantages and disadvantages of the relevant applications, are also analyzed and discussed in depth, targeting to summarize the methodology and provide design reference for the research and development of radio frequency front-end integrated power amplifier chips for modern wireless communication systems. Eventually, the development trend of radio frequency/microwave power amplifier chip technology and the industry direction are outlooked.
Based on a comprehensive review and classification of the concepts, types, and realization processes of radio frequency/microwave power amplifier chips, the research status and urgent technical problems of key techniques such as high frequency, linearity improvement, energy conversion efficiency improvement, bandwidth expansion and highly integrated packaging and so on, are focused on. The mainstream realization of each key technology, typical cases of research and development, as well as the advantages and disadvantages of the relevant applications, are also analyzed and discussed in depth, targeting to summarize the methodology and provide design reference for the research and development of radio frequency front-end integrated power amplifier chips for modern wireless communication systems. Eventually, the development trend of radio frequency/microwave power amplifier chip technology and the industry direction are outlooked.
, Available online , doi: 10.12178/1001-0548.2023242
Abstract:
To assist in monitoring occasional processing errors in the precision machining of Potassium Dihydrogen Phosphate (KDP) on ultra-precision fly-cutting machines, this paper combines key feature extraction of vibration data and temperature data during the machining process to establish a predictive model for crystal processing surfaces. Based on ResNet-18, the relationship between vibration data and whether the KDP crystal surface is qualified is analyzed, and binary classification predictions are made. The model achieves an accuracy of 88.5% on the test set. Meanwhile, based on the XGBoost model, the relationship between temperature data and the low-frequency index P-V of KDP crystal surface quality is analyzed and predicted. The experimental results show that the prediction model can predict the surface quality of the processed element quickly, and the overall error is within an acceptable range. By analyzing the processing errors, a complete machine tool model is constructed. The transient temperature field of the machine tool under long-time processing is calculated using finite element analysis. The simulation results show that the maximum temperature of the machine tool reaches 26.9°C after 8580 s of the operation. Experimental verification confirms the accuracy of the simulation results and supports the conclusion that the "decline in KDP crystal processing quality in the later stage" is related to "the continuous warming of the machine tool spindle system during the processing".
To assist in monitoring occasional processing errors in the precision machining of Potassium Dihydrogen Phosphate (KDP) on ultra-precision fly-cutting machines, this paper combines key feature extraction of vibration data and temperature data during the machining process to establish a predictive model for crystal processing surfaces. Based on ResNet-18, the relationship between vibration data and whether the KDP crystal surface is qualified is analyzed, and binary classification predictions are made. The model achieves an accuracy of 88.5% on the test set. Meanwhile, based on the XGBoost model, the relationship between temperature data and the low-frequency index P-V of KDP crystal surface quality is analyzed and predicted. The experimental results show that the prediction model can predict the surface quality of the processed element quickly, and the overall error is within an acceptable range. By analyzing the processing errors, a complete machine tool model is constructed. The transient temperature field of the machine tool under long-time processing is calculated using finite element analysis. The simulation results show that the maximum temperature of the machine tool reaches 26.9°C after 8580 s of the operation. Experimental verification confirms the accuracy of the simulation results and supports the conclusion that the "decline in KDP crystal processing quality in the later stage" is related to "the continuous warming of the machine tool spindle system during the processing".
, Available online , doi: 10.12178/1001-0548.2023229
Abstract:
The characteristics of different types of social robots on large online social platforms such as Twitter, Facebook and Sina Weibo were analyzed. Based on the social robot detection framework, the advantages and disadvantages and applicability of social robot detection models based on machine learning, deep learning and other emerging detection methods were summarized and analyzed. It is found that social robots with different platforms and attack targets need to extract multi-dimensional features and design corresponding detection methods. Finally, this paper deeply explores and analyzes how to reduce the harm of social robots and measures to cope with the challenges of coexistence between human and social robots, and discusses and looks forward to how to improve the recognition accuracy and the development of hot technologies.
The characteristics of different types of social robots on large online social platforms such as Twitter, Facebook and Sina Weibo were analyzed. Based on the social robot detection framework, the advantages and disadvantages and applicability of social robot detection models based on machine learning, deep learning and other emerging detection methods were summarized and analyzed. It is found that social robots with different platforms and attack targets need to extract multi-dimensional features and design corresponding detection methods. Finally, this paper deeply explores and analyzes how to reduce the harm of social robots and measures to cope with the challenges of coexistence between human and social robots, and discusses and looks forward to how to improve the recognition accuracy and the development of hot technologies.
, Available online , doi: 10.12178/1001-0548.2023170
Abstract:
A large amount of images are streaming into the internet and the online image retrieval task become more and more popular. To guarantee the online image retrieval performance, online hashing algorithms are utilized to re-learn the hash functions and re-generate the hash codes of the new and old samples in real time. As time went by, the amount of old dataset is very large, and the time complexity of re-generating hash codes become unacceptable. To avoid the above problems, we propose a novel Asymmetric Deep Online Hashing (ADOH) which trains a deep online hashing network in an asymmetric manner. To improve the online retrieval efficiency, ADOH only generates the hash code of the new samples and do not update the old samples’ hash codes. During the training process, ADOH minimizes the difference between the hash codes’ inner product and the similarity matrix, which can preserve the pair-wise semantic similarity relationship. Moreover, ADOH establishes the classification loss and the label embedding model to learn the samples’ semantic information, which makes the generated hash codes more semantically discriminative. We conduct the approximate nearest neighbor retrieval comparative experiments on three widely used datasets including the cifar-10 dataset, the mnist dataset and the Places205 dataset. The results show that the online approximate nearest neighbor retrieval performance of ADOH outperforms the other 8 existing online hashing methods.
A large amount of images are streaming into the internet and the online image retrieval task become more and more popular. To guarantee the online image retrieval performance, online hashing algorithms are utilized to re-learn the hash functions and re-generate the hash codes of the new and old samples in real time. As time went by, the amount of old dataset is very large, and the time complexity of re-generating hash codes become unacceptable. To avoid the above problems, we propose a novel Asymmetric Deep Online Hashing (ADOH) which trains a deep online hashing network in an asymmetric manner. To improve the online retrieval efficiency, ADOH only generates the hash code of the new samples and do not update the old samples’ hash codes. During the training process, ADOH minimizes the difference between the hash codes’ inner product and the similarity matrix, which can preserve the pair-wise semantic similarity relationship. Moreover, ADOH establishes the classification loss and the label embedding model to learn the samples’ semantic information, which makes the generated hash codes more semantically discriminative. We conduct the approximate nearest neighbor retrieval comparative experiments on three widely used datasets including the cifar-10 dataset, the mnist dataset and the Places205 dataset. The results show that the online approximate nearest neighbor retrieval performance of ADOH outperforms the other 8 existing online hashing methods.
, Available online , doi: 10.12178/1001-0548.2023213
Abstract:
The LLC resonant converter has significant merits, such as high conversion efficiency and soft switching characteristics. However, in the field of DC charging, the wide output voltage gain requires the operating frequency of the LLC resonant converter must cover a wide range, which cause the loss of the conversion efficiency, the complex design of magnetic components, and limited light-load regulation, etc. In this paper, an LCL-T/LLC resonant converter with a multiplexing resonant inductor for the on-board charging of electric vehicles is proposed. On the secondary side of the transformer, the resonant inductor is operating with the LCL-T resonant tank for the constant current(CC) charge of the battery. The inherent characteristic of the LCL-T can limit the range of the operating frequency. While for the constant voltage(CV) charge, the resonant tank is modified by the auxiliary branch, and the resonant inductor is operating in the LLC converter, which works as a CV source. Due to the resonant inductor is placed on the secondary side of the transformer, the core loss during the dead time is reduced and the conversion efficiency is improved. The operation mode and the DC gain characteristics of the proposed converter are analyzed in detail. Moreover, the design precede of key parameters is provided. Finally, a 3.3kW, 400V/ 330V experimental prototype is built to verify the feasibility and effectiveness of the proposed resonant converter and its control method.
The LLC resonant converter has significant merits, such as high conversion efficiency and soft switching characteristics. However, in the field of DC charging, the wide output voltage gain requires the operating frequency of the LLC resonant converter must cover a wide range, which cause the loss of the conversion efficiency, the complex design of magnetic components, and limited light-load regulation, etc. In this paper, an LCL-T/LLC resonant converter with a multiplexing resonant inductor for the on-board charging of electric vehicles is proposed. On the secondary side of the transformer, the resonant inductor is operating with the LCL-T resonant tank for the constant current(CC) charge of the battery. The inherent characteristic of the LCL-T can limit the range of the operating frequency. While for the constant voltage(CV) charge, the resonant tank is modified by the auxiliary branch, and the resonant inductor is operating in the LLC converter, which works as a CV source. Due to the resonant inductor is placed on the secondary side of the transformer, the core loss during the dead time is reduced and the conversion efficiency is improved. The operation mode and the DC gain characteristics of the proposed converter are analyzed in detail. Moreover, the design precede of key parameters is provided. Finally, a 3.3kW, 400V/ 330V experimental prototype is built to verify the feasibility and effectiveness of the proposed resonant converter and its control method.
