检索结果-内蒙古大学图书馆

Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)

作者： Rei Hamakawa Michiharu Niimi Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan

ISBN: (数字)9798350367331

ISBN: (纸本)9798350367348

This paper first provides an overview of the English pronunciation learning support tool. The tool aims to use "accent-modified speech that retains the learner’s voice quality" as the "target speech." We propose a new conversion model based on conventional methods for this accent conversion. Specifically, we improve the conventional LSTM-based DNN model for accent conversion by adopting a transformer-based model. Our experiments investigated the model’s ability to handle the unique katakana pronunciation characteristic of Japanese speakers. The results confirmed the effectiveness of the proposed conversion method, although challenges remain, such as the scarcity of Japanese speech data and the need to improve the accuracy of speaker identity retention.

关键词： Accuracy Synthesizers Training data Symbols Switches Information processing Phonetics Transformers Data models Speech processing

来源：评论

学校读者我要写书评

暂无评论

Optimal Resource Allocation for NOMA Wireless Networks

引用

computers, Materials & Continua 2023年第2期74卷 3249-3261页

作者： Fahad R.Albogamy M.A.Aiyashi Fazirul Hisyam Hashim Imran Khan Bong Jun Choi Turabah University College Computer Sciences ProgramTaif UniversityP.O.Box 11099Taif21944Saudi Arabia Department of Mathematics Jazan UniversityJazanP.O.Box 218Saudi Arabia Department of Computer and Communication Systems Engineering Faculty of EngineeringUniversiti Putra Malaysia(UPM)43400SerdangMalaysia Department of Electrical Engineering University of Engineering and Technology PeshawarP.O.Box 814Pakistan School of Computer Science and Engineering Soongsil UniversitySeoulKorea

The non-orthogonal multiple access(NOMA)method is a novel multiple access technique that aims to increase spectral efficiency(SE)and accommodate enormous user ***-user signals are superimposed and transmitted in the power domain at the transmitting end by actively implementing controllable interference information,and multi-user detection algorithms,such as successive interference cancellation(SIC),are performed at the receiving end to demodulate the necessary user *** its basic signal waveform,like LTE baseline,could be based on orthogonal frequency division multiple access(OFDMA)or discrete Fourier transform(DFT)-spread OFDM,NOMA superimposes numerous users in the power *** contrast to the orthogonal transmission method,the nonorthogonal method can achieve higher spectrum ***,it will increase the complexity of its *** power allocation techniques will have a direct impact on the system’s *** a result,in order to boost the system capacity,an efficient power allocation mechanism must be *** research developed an efficient technique based on conjugate gradient to solve the problem of downlink power *** major goal is to maximize the users’maximum weighted sum *** suggested algorithm’s most notable feature is that it converges to the global optimal *** compared to existing methods,simulation results reveal that the suggested technique has a better power allocation capability.

关键词： Non-orthogonalmultiple access wireless networks power domain 5G networks

来源：评论

学校读者我要写书评

暂无评论

Exploring the Effect of Activation Function on Transformer Model Performance for Official Announcement Translator from Indonesian to Sundanese Languages

Exploring the Effect of Activation Function on Transformer M...

引用

2023 International Conference on computer science, Information Technology and engineering, ICCoSITE 2023

作者： Wijanarko, Bambang Dwi Fitria Murad, Dina Heryadi, Yaya Tho, Cuk Hashimoto, Kiyota BINUS Online Learning Computer Science Department Jakarta11480 Indonesia BINUS Online Learning Information Systems Department Jakarta11480 Indonesia BINUS Graduate Program-Doctor of Computer Science Computer Science Department Jakarta11480 Indonesia Prince of Songkla University Interdisciplinary Graduate School or Earth System Science and Andaman Natural Disaster Management Phuket Campus Phuket83120 Thailand

ISBN: (纸本)9798350320954

Automated language translation involving low-resource language has gained wide interest from many research communities in the past decade. One lesson learned from the past COVID-19 pandemic, particularly in Indonesia, is that many local Governments have to release regular public announcements to keep people following health protocol especially when they are in public areas. Many studies showed some evidence that rural people in Indonesia which covers a large proportion of Indonesia's population, feel more convenience receiving official announcements in their local language. However, translating official announcement from the national language to many local languages in Indonesia require many experienced bilingual translators and time. This paper presents exploration results in developing an automated language translator model to translate texts in Bahasa Indonesia to the Sundanese language. In particular, this study aims to explore the effect of ReLU, Sigmoid, and Tanh activation functions of the Vanilla Transformer Model on its translation performance. The experiment results showed that the activation function under study gives similar training accuracy (0.98). However, ReLU achieves better performance than Tanh in terms of validation accuracy, training loss, and validation loss. © 2023 IEEE.

关键词： Chemical activation

来源：评论

学校读者我要写书评

暂无评论

A Study on DDoS Attacks Detection on IoT Devices Using Machine Learning for Microcontrollers

A Study on DDoS Attacks Detection on IoT Devices Using Machi...

引用

IEEE International Conference on Consumer Electronics (ICCE)

作者： Jedidah Mwaura Shunsuke Araki Ken’Ichi Kakizaki Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology FuKuoka Japan Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Fukuoka Japan

In recent years, there has been a proliferation of Internet of Things (IoT) devices, and so has been the attacks on them. In this paper we will propose a methodology to detect Distributed Denial of Service (DDoS) attacks on IoT devices using Machine Learning for Microcontrollers. We will discuss a model which we made for Arduino Nano 33 BLE Sense using Machine Learning for Microcontrollers. Additionally, we will discuss results of our proposal in detecting DDoS attacks on IoT devices. Lastly, we will describe the feasibility of our model on IoT devices.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Generation of Photo Slideshow with Song based on Closeness between Concept of Lyrics and That of Images

Generation of Photo Slideshow with Song based on Closeness b...

引用

Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)

作者： Mei Hashimoto Michiharu Niimi Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan

ISBN: (数字)9798350367331

ISBN: (纸本)9798350367348

This paper proposes a method that allows users to easily convert a large number of still images into movies by displaying photos in sync with memories or favorite songs, which offers a new media viewing method for users to look back at the vast number of photos in their photo folders. We assume here that users select a song and have images stored in local PC. One method of synchronization involves selecting images based on elapsed time of song. However, since lyrics convey meaning, it can be better to display the images that align with both lyrics’ meaning and images’ meaning matched. In order to do this, we need the space which shares the concept of both lyrics and images, and select colors as the concept. Displayed images are determined based on the distance between points that are mapped from words and images to color space. Specifically, the color concepts from words are extracted using the Color Image Scale, and that from images are determined as the most frequent colors by cluster analysis. Experimental results show that it is possible to easily create movies where images with colors that match the mood of the lyrics are displayed.

关键词： Image color analysis Mood Refining Asia Information processing Color Media Motion pictures Synchronization

来源：评论

学校读者我要写书评

暂无评论

Global-Local Consistency for Cost-Effective Anomaly Localization with Point Annotations

Global-Local Consistency for Cost-Effective Anomaly Localiza...

引用

IIAI International Conference on Advanced Applied Informatics (IIAIAAI)

作者： Keiichi Nakanishi Ryoya Katafuchi Terumasa Tokunaga Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan

ISBN: (数字)9798350377903

ISBN: (纸本)9798350377910

Creating pixel-level ground-truth (GT) masks is quite costly for deep learning-based image segmentation. Specialists in areas such as anomaly detection and medical diagnostics face difficulties in producing many GT masks due to limited resources. To reduce this burden, we propose a cost-effective image segmentation framework with point annotations (CoSPA) that performs image segmentation with only point annotations. Point annotations refer to the partial and sparse labeled coordinates in an image. The key idea is to ensure consistency between the predictions for the same coordinates from the two different networks of CoSPA. This new consistency enables the CoSPA to improve segmentation performance and extend to semi-supervised learning. For the MVTec AD dataset, we verified the cost-effectiveness of CoSPA through an anomaly detection task. We demonstrated that the point annotating costs were reduced by 80% compared to creating GT masks. Subsequently, the CoSPA realized by 87% of the mean Intersection over Union (mIoU) achieved using the fully supervised method, DeepLabV3+. Moreover, the mIoU of CoSPA using only 30% of all point annotations defeated that of the unsupervised method, PaDiM. This study offers a new direction for economic anomaly localization.

关键词： Location awareness Image segmentation Costs Annotations Supervised learning Semisupervised learning Medical diagnosis Informatics Anomaly detection Faces

来源：评论

学校读者我要写书评

暂无评论

Framework for Integration of Generative AI into Metaverse Asset Creation

Framework for Integration of Generative AI into Metaverse As...

引用

Intelligent Metaverse Technologies & Applications (iMETA), International Conference on

作者： Rio Mukhtarom Paweroi Mario Köppen Department of Computer Science and Networks Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan

ISBN: (数字)9798350351514

ISBN: (纸本)9798350351521

Metaverse, a virtual world, is developing rapidly and is widely used in multi-sector. The number of users is projected to increase year over year. Due to the development of the metaverse platform, digital asset creation is demanding. Creating high-quality and diverse 3D digital objects is challenging. This study proposes the frameworks for integrating generative AI to create diverse 3D assets into the metaverse. We study different approaches for asset creation, i.e., generative 3D model-based, generative image projection-based, and generative language script-based. Creators can use this workflow to optimize the creation of 3D assets. Moreover, this study compares the results of generative AI and procedural generation on generating diverse 3D objects. The result shows that generative AI can simplify 3D creation and generate more diverse objects.

关键词： Solid modeling Three-dimensional displays Procedural generation Generative AI Metaverse Complexity theory Usability

来源：评论

学校读者我要写书评

暂无评论

Detection of Diffusion-Generated Images Using Sparse Coding

Detection of Diffusion-Generated Images Using Sparse Coding

引用

Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)

作者： Daishi Tanaka Michiharu Niimi Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Fukuoka Japan Faculty of Computer Science and Systems Engineering Kyushu Institute of Technology Fukuoka Japan

ISBN: (数字)9798350367331

ISBN: (纸本)9798350367348

This paper proposes a method for detecting images generated by diffusion models using sparse coding. In the diffusion model, an image can be generated by removing noise from a noisy image. This different generation process from real images leads us to believe that there may be a statistical difference in pixels between the real and the generated images. Specifically, the image is divided into small patch regions, and all patch images are reconstructed using the basis image. In this process, sparse coefficients that contain many zeros are obtained using sparse coding, and features are calculated from the obtained coefficients. Then, a simple discriminator using the features as input is trained with a small number of data to discriminate the diffusion-generated images. In our experiments, we evaluated the proposed method on six datasets created using three diffusion models and two real image datasets. Experiments were also conducted to evaluate the robustness of the proposed method against JPEG compression. Experimental results show that our proposed method is sufficiently robust against JPEG compression with as few as 1800 training data.

关键词： Image coding Noise Transform coding Training data Information processing Diffusion models Feature extraction Robustness Noise measurement Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Unleashing the Potential of Reinforcement Learning for Enhanced Personalized Education 53

Unleashing the Potential of Reinforcement Learning for Enhan...

引用

53rd IEEE ASEE Frontiers in Education International Conference, FIE 2023

作者： Fernandes, Chelsea William Miari, Tahereh Rafatirad, Setareh Sayadi, Hossein California State University Department of Computer Engineering and Computer Science Long BeachCA United States Center for Information Systems and Technology Claremont Graduate University CA United States University of California Department of Computer Science DavisCA United States

ISBN: (纸本)9798350336429

Providing personalized content that meets the diverse needs of learners is a significant challenge for today's learning systems. Current educational platforms lack the capability to effectively support the needs of learners with varying intellectual abilities, learning pace, preferences, and academic backgrounds. To overcome this challenge, there is a pressing need for sustainable educational tools that can adapt to the individual needs of student learners. Our study proposes Reinforced-EDU, an intelligent and effective reinforcement learning-guided framework based on Q-Learning technique. The framework adaptively schedules course assignments and educational activities based on students' characteristics, preferences, skills, and academic backgrounds. It considers different student-centered academic factors to prescribe a suitable learning plan that maximizes the students' overall grade and satisfaction rate while reducing dropouts. We identify four distinct learning modes tailored towards effective logical reasoning that align best with the students' preferences and characteristics. Our proposed intelligent method, Reinforced-EDU, can dynamically assign an optimal learning path with 95% accuracy based on students' preferences and pre-knowledge. © 2023 IEEE.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Deep Semantic and Attentive Network for Unsupervised Video Summarization

引用

ACM Transactions on Multimedia Computing, Communications and Applications 2022年第2期18卷 1-21页

作者： Zhong, Sheng-Hua Lin, Jingxu Lu, Jianglin Fares, Ahmed Ren, Tongwei College of Computer Science and Software Engineering Shenzhen University Shenzhen518060 China Department of Electrical Engineering Computer Systems Engineering Program Faculty of Engineering at Shoubra Benha University Cairo Egypt State Key Laboratory for Novel Software Technology Nanjing University Nanjing210023 China

With the rapid growth of video data, video summarization is a promising approach to shorten a lengthy video into a compact version. Although supervised summarization approaches have achieved state-of-the-art performance, they require frame-level annotated labels. Such an annotation process is time-consuming and tedious. In this article, we propose a novel deep summarization framework named Deep Semantic and Attentive Network for Video Summarization (DSAVS) that can select the most semantically representative summary by minimizing the distance between video representation and text representation without any frame-level labels. Another challenge associated with video summarization tasks mainly originates from the difficulty of considering temporal information over a long time. Long Short-Term Memory (LSTM) performs well for temporal dependencies modeling but does not work well with long video clips. Therefore, we introduce a self-attention mechanism into our summarization framework to capture the long-range temporal dependencies among the frames. Extensive experiments on two popular benchmark datasets, i.e., SumMe and TVSum, show that our proposed framework outperforms other state-of-the-art unsupervised approaches and even most supervised methods. © 2022 Association for Computing Machinery.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：