The challenging task of handwriting style synthesis requires capturing the individuality and diversity of human handwriting.The majority of currently available methods use either a generative adversarial network(GAN)o...The challenging task of handwriting style synthesis requires capturing the individuality and diversity of human handwriting.The majority of currently available methods use either a generative adversarial network(GAN)or a recurrent neural network(RNN)to generate new handwriting styles.This is why these techniques frequently fall short of producing diverse and realistic text pictures,particularly for terms that are not commonly used.To resolve that,this research proposes a novel deep learning model that consists of a style encoder and a text generator to synthesize different handwriting styles.This network excels in generating conditional text by extracting style vectors from a series of style images.The model performs admirably on a range of handwriting synthesis tasks,including the production of text that is out-of-vocabulary.It works more effectively than previous approaches by displaying lower values on key Generative Adversarial Network evaluation metrics,such Geometric Score(GS)(3.21×10^(-5))and Fréchet Inception Distance(FID)(8.75),as well as text recognition metrics,like Character Error Rate(CER)and Word Error Rate(WER).A thorough component analysis revealed the steady improvement in image production quality,highlighting the importance of specific handwriting styles.Applicable fields include digital forensics,creative writing,and document security.展开更多
最近混淆网络在融合多个机器翻译结果中展示很好的性能.然而为了克服在不同的翻译系统中不同的词序,假设对齐在混淆网络的构建上仍然是一个重要的问题.但以往的对齐方法都没有考虑到语义信息.本文为了更好地改进系统融合的性能,提出了...最近混淆网络在融合多个机器翻译结果中展示很好的性能.然而为了克服在不同的翻译系统中不同的词序,假设对齐在混淆网络的构建上仍然是一个重要的问题.但以往的对齐方法都没有考虑到语义信息.本文为了更好地改进系统融合的性能,提出了用词义消歧(Word sense disambiguation,WSD)来指导混淆网络中的对齐.同时骨架翻译的选择也是通过计算句子间的相似度来获得的,句子的相似性计算使用了二分图的最大匹配算法.为了使得基于WordNet词义消歧方法融入到系统中,本文将翻译错误率(Translation error rate,TER)算法进行了改进,实验结果显示本方法的性能好于经典的TER算法的性能.展开更多
Efficient anti-jamming rateless coding based on cognitive Orthogonal Frequency Division Multiplexing (OFDM) modulation in Cognitive Radio Network (CRN) is mainly discussed. Rateless coding with small redundancy and lo...Efficient anti-jamming rateless coding based on cognitive Orthogonal Frequency Division Multiplexing (OFDM) modulation in Cognitive Radio Network (CRN) is mainly discussed. Rateless coding with small redundancy and low complexity is presented, and the optimal design methods of building rateless codes are also proposed. In CRN, anti-jamming rateless coding could recover the lost packets in parallel channels of cognitive OFDM, thus it protects Secondary Users (SUs) from the in-terference by Primary Users (PUs) efficiently. Frame Error Rate (FER) and throughput performance of SU employing anti-jamming rateless coding are analyzed in detail. Performance comparison between rateless coding and piecewise coding are also presented. It is shown that, anti-jamming rateless coding provides low FER and Word Error Rate (WER) performance with uniform sub-channel selection. Meanwhile, it is also verified that, in higher jamming rate and longer code redundancy scenario, rateless coding method could achieve better FER and throughput performance than another anti-jamming coding schemes.展开更多
当向机器翻译模型输入序列时,随着序列长度的不断增长,会出现长距离约束即输入输出序列的长度被限制在固定范围内的问题,因此所建模型的能力会受到约束。序列到序列模型(sequence to sequence model)可以解决长距离约束问题,但单纯的序...当向机器翻译模型输入序列时,随着序列长度的不断增长,会出现长距离约束即输入输出序列的长度被限制在固定范围内的问题,因此所建模型的能力会受到约束。序列到序列模型(sequence to sequence model)可以解决长距离约束问题,但单纯的序列到序列模型无法对翻译中要参考词语前后或其他位置的内容来改善翻译质量的行为进行建模。为了弥补该缺陷,提出了注意力机制(attention mechanism)。针对以上问题,报告了机器翻译及部分模型的研究现状,简述了深度学习框架,分析了基于神经网络的机器翻译及注意力机制原理,并对使用PyTorch实现的序列到序列模型及注意力机制进行了研究,通过分析翻译的时间消耗和翻译后的词错率以及评价标准的值来评价模型。最终该模型在英法数据集上取得了一定的效果。展开更多
Communication is a significant part of being human and living in the world.Diverse kinds of languages and their variations are there;thus,one person can speak any language and cannot effectively communicate with one w...Communication is a significant part of being human and living in the world.Diverse kinds of languages and their variations are there;thus,one person can speak any language and cannot effectively communicate with one who speaks that language in a different accent.Numerous application fields such as education,mobility,smart systems,security,and health care systems utilize the speech or voice recognition models abundantly.Though,various studies are focused on the Arabic or Asian and English languages by ignoring other significant languages like Marathi that leads to the broader research motivations in regional languages.It is necessary to understand the speech recognition field,in which the major concentrated stages are feature extraction and classification.This paper emphasis developing a Speech Recognition model for the Marathi language by optimizing Recurrent Neural Network(RNN).Here,the preprocessing of the input signal is performed by smoothing and median filtering.After preprocessing the feature extraction is carried out using MFCC and Spectral features to get precise features from the input Marathi Speech corpus.The optimized RNN classifier is used for speech recognition after completing the feature extraction task,where the optimization of hidden neurons in RNN is performed by the Grasshopper Optimization Algorithm(GOA).Finally,the comparison with the conventional techniques has shown that the proposed model outperforms most competing models on a benchmark dataset.展开更多
Energy efficiency （EE） can be enhanced by retransmissions and combining in hybrid automatic repeat request （HARQ） system. However, it is difficult to optimize the transmit power of each retransmission when the acc...Energy efficiency （EE） can be enhanced by retransmissions and combining in hybrid automatic repeat request （HARQ） system. However, it is difficult to optimize the transmit power of each retransmission when the accurate retransmission number and future channel state information （CSI） cannot be obtained. This paper proposes a simple energy efficient HARQ scheme for point-to-point wireless communication. In the proposed scheme, the conditional word error rate （WER） of each retransmission is fixed and the transmit power is adapted correspondingly. Three performance metrics are analyzed including average transmission number, throughput and EE. Compared with the conventional equal power HARQ scheme, the proposed scheme can significantly improve the EE and other two metrics under the same constraint of average transmit power or average energy consumption. Furthermore, it is found that, selecting a conditional WER which is slightly smaller than the optimal one is sufficient for practical implementation.展开更多
基金supported by the National Research Foundation of Korea(NRF)Grant funded by the Korean government(MSIT)(NRF-2023R1A2C1005950).
文摘The challenging task of handwriting style synthesis requires capturing the individuality and diversity of human handwriting.The majority of currently available methods use either a generative adversarial network(GAN)or a recurrent neural network(RNN)to generate new handwriting styles.This is why these techniques frequently fall short of producing diverse and realistic text pictures,particularly for terms that are not commonly used.To resolve that,this research proposes a novel deep learning model that consists of a style encoder and a text generator to synthesize different handwriting styles.This network excels in generating conditional text by extracting style vectors from a series of style images.The model performs admirably on a range of handwriting synthesis tasks,including the production of text that is out-of-vocabulary.It works more effectively than previous approaches by displaying lower values on key Generative Adversarial Network evaluation metrics,such Geometric Score(GS)(3.21×10^(-5))and Fréchet Inception Distance(FID)(8.75),as well as text recognition metrics,like Character Error Rate(CER)and Word Error Rate(WER).A thorough component analysis revealed the steady improvement in image production quality,highlighting the importance of specific handwriting styles.Applicable fields include digital forensics,creative writing,and document security.
文摘最近混淆网络在融合多个机器翻译结果中展示很好的性能.然而为了克服在不同的翻译系统中不同的词序,假设对齐在混淆网络的构建上仍然是一个重要的问题.但以往的对齐方法都没有考虑到语义信息.本文为了更好地改进系统融合的性能,提出了用词义消歧(Word sense disambiguation,WSD)来指导混淆网络中的对齐.同时骨架翻译的选择也是通过计算句子间的相似度来获得的,句子的相似性计算使用了二分图的最大匹配算法.为了使得基于WordNet词义消歧方法融入到系统中,本文将翻译错误率(Translation error rate,TER)算法进行了改进,实验结果显示本方法的性能好于经典的TER算法的性能.
基金Supported by the National Natural Science Foundation of China (No. 60972039)the Scientific Planning Project of Zhejiang Province entitled "Research and Development of Smart Antenna for the Next Generation Mobile Com-munications Based on TDD"the Young Staff Startup Research Foundation of Hangzhou Dianzi University entitled "Research on Key Technologies of Resource Allocation in Cognitive Radio Networks Based on Multicarrier Modulation"
文摘Efficient anti-jamming rateless coding based on cognitive Orthogonal Frequency Division Multiplexing (OFDM) modulation in Cognitive Radio Network (CRN) is mainly discussed. Rateless coding with small redundancy and low complexity is presented, and the optimal design methods of building rateless codes are also proposed. In CRN, anti-jamming rateless coding could recover the lost packets in parallel channels of cognitive OFDM, thus it protects Secondary Users (SUs) from the in-terference by Primary Users (PUs) efficiently. Frame Error Rate (FER) and throughput performance of SU employing anti-jamming rateless coding are analyzed in detail. Performance comparison between rateless coding and piecewise coding are also presented. It is shown that, anti-jamming rateless coding provides low FER and Word Error Rate (WER) performance with uniform sub-channel selection. Meanwhile, it is also verified that, in higher jamming rate and longer code redundancy scenario, rateless coding method could achieve better FER and throughput performance than another anti-jamming coding schemes.
文摘当向机器翻译模型输入序列时,随着序列长度的不断增长,会出现长距离约束即输入输出序列的长度被限制在固定范围内的问题,因此所建模型的能力会受到约束。序列到序列模型(sequence to sequence model)可以解决长距离约束问题,但单纯的序列到序列模型无法对翻译中要参考词语前后或其他位置的内容来改善翻译质量的行为进行建模。为了弥补该缺陷,提出了注意力机制(attention mechanism)。针对以上问题,报告了机器翻译及部分模型的研究现状,简述了深度学习框架,分析了基于神经网络的机器翻译及注意力机制原理,并对使用PyTorch实现的序列到序列模型及注意力机制进行了研究,通过分析翻译的时间消耗和翻译后的词错率以及评价标准的值来评价模型。最终该模型在英法数据集上取得了一定的效果。
基金Taif University Researchers Supporting Project number(TURSP-2020/349),Taif University,Taif,Saudi Arabia.
文摘Communication is a significant part of being human and living in the world.Diverse kinds of languages and their variations are there;thus,one person can speak any language and cannot effectively communicate with one who speaks that language in a different accent.Numerous application fields such as education,mobility,smart systems,security,and health care systems utilize the speech or voice recognition models abundantly.Though,various studies are focused on the Arabic or Asian and English languages by ignoring other significant languages like Marathi that leads to the broader research motivations in regional languages.It is necessary to understand the speech recognition field,in which the major concentrated stages are feature extraction and classification.This paper emphasis developing a Speech Recognition model for the Marathi language by optimizing Recurrent Neural Network(RNN).Here,the preprocessing of the input signal is performed by smoothing and median filtering.After preprocessing the feature extraction is carried out using MFCC and Spectral features to get precise features from the input Marathi Speech corpus.The optimized RNN classifier is used for speech recognition after completing the feature extraction task,where the optimization of hidden neurons in RNN is performed by the Grasshopper Optimization Algorithm(GOA).Finally,the comparison with the conventional techniques has shown that the proposed model outperforms most competing models on a benchmark dataset.
基金supported by National Natural Science Foundation of China (61072059)
文摘Energy efficiency （EE） can be enhanced by retransmissions and combining in hybrid automatic repeat request （HARQ） system. However, it is difficult to optimize the transmit power of each retransmission when the accurate retransmission number and future channel state information （CSI） cannot be obtained. This paper proposes a simple energy efficient HARQ scheme for point-to-point wireless communication. In the proposed scheme, the conditional word error rate （WER） of each retransmission is fixed and the transmit power is adapted correspondingly. Three performance metrics are analyzed including average transmission number, throughput and EE. Compared with the conventional equal power HARQ scheme, the proposed scheme can significantly improve the EE and other two metrics under the same constraint of average transmit power or average energy consumption. Furthermore, it is found that, selecting a conditional WER which is slightly smaller than the optimal one is sufficient for practical implementation.
