检索结果-内蒙古大学图书馆

Enhancing Automatic Speech Recognition for Punjabi Dialects: An Experimental Analysis of Incorporating Prosodic Features and Acoustic Variability Mitigation

引用

SN computer science 2024年第6期5卷 747页

作者： Bhardwaj, Vivek Gera, Tanya Thakur, Deepak Singh, Amitoj School of Computer Science and Engineering Manipal University Jaipur Chitkara University Institute of Engineering and Technology Chitkara University Jagat Guru Nanak Dev Punjab State Open University

The development of Automatic Speech Recognition (ASR) systems has varied significantly across the roughly 6500 languages that make up the world's spoken languages. There is a lack of effective ASR systems due to the unavailability of essential resources including high-quality speech corpora, pronunciation dictionaries, and transcriptions, for the low resource languages including the Punjabi language with its several dialects. With a focus on the Malwai, Doabi, Majhi, and Puadhi dialects, this work initiates an experimental investigation of ASR systems developed exclusively for the Punjabi language. To improve speech recognition rates, the study offers a novel method that combines prosodic variables like pitch and probability of voicing (POV) with Mel-frequency cepstral coefficients (MFCC). The goal of this combination of elements is to accurately reproduce the tonal qualities of the Punjabi language. The work addresses the issue of inter-speaker acoustic variabilities using approaches like speaker adaptive training (SAT) and vocal-tract length normalisation (VTLN). The word error rate (WER), which reduces upto 9.25% for the different Punjabi dialects, has been shown to be significantly improvements because of the proposed hybridization of prosodic characteristics. This result highlights how this method may improve ASR systems for tonal languages with limited resources. In addition to demonstrating the value of adding prosodic features and minimising inter-speaker variations, the study advances the field by describing a novel paradigm for creating ASR systems for languages with constrained and sustainable use of resources. This study also presents present state of art as well as emerging trends and a few open innovative problems for research community. Future researchers may carefully observe literature gap to apply this strategy to other low-resource languages and dialects, further demonstrating the effectiveness of the suggested approach. © The Author(s), under exc

关键词： Acoustic Automatic speech recognition Pitch Punjabi dialects

来源：评论

学校读者我要写书评

暂无评论

Idecomp: imbalance-aware decomposition for class-decomposed classification using conditional GANs

引用

Discover Artificial Intelligence 2023年第1期3卷 31页

作者： Buczek, Patryk Zidan, Usama Gaber, Mohamed Medhat Abdelsamea, Mohammed M. School of Computing and Digital Technology Birmingham City University Birmingham100190 United Kingdom Faculty of Computer Science and Engineering Galala University Suez435611 Egypt Department of Computer Science Faculty of Computers and Information Assiut University Assiut71515 Egypt

Medical image classification tasks frequently encounter challenges associated with class imbalance, resulting in biased model training and suboptimal classification performance. To address this issue, the combination of class decomposition and transfer learning has proven to be effective in classifying imbalanced medical imaging datasets. Nevertheless, in order to further augment the performance gains achieved through the utilisation of class decomposition within deep learning frameworks, we propose a novel model coined imbalance-Aware Decomposition for Class-Decomposed Classification (iDeComp) model. By incorporating a conditional Generative Adversarial Network (GAN) model, iDeComp is capable of generating additional samples specifically tailored to underrepresented decomposed subclasses. This paper investigates the application of iDeComp using two different medical imaging datasets. iDeComp selects underrepresented samples from the training set of the sublevel classes within each dataset, which are then employed to train separate conditional Deep Convolutional GAN (DCGAN) models and verification models. The conditional DCGAN model is responsible for generating additional samples, while the verification model critically evaluates the appropriateness of the synthesised images. Subsequently, the resulting augmented samples are utilized to train the classification model. To assess the effectiveness of iDeComp, we employ various evaluation metrics including accuracy, precision, recall, and F1 score. The results obtained from our experiments clearly indicate that iDeComp outperforms existing approaches in terms of classifying both imbalanced datasets. © The Author(s) 2023.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Customer Segmentation of E-commerce data using K-means Clustering Algorithm 13

Customer Segmentation of E-commerce data using K-means Clust...

引用

13th International Conference on Cloud Computing, Data science and engineering, Confluence 2023

作者： Rajput, Lucky Singh, Shailendra Narayan Amity University Amity School of Engineering and Technology Department of Computer Science and Engineering Noida India

ISBN: (纸本)9781665462631

Several data mining techniques, such as classification, clustering, regression, etc., are used to determine the purchasing behaviour of customers to create value for money in businesses. In this paper, clustering is implemented on a real-time data set of an e-commerce firm that aims to decide whether to focus on its website or mobile application. The K-means clustering algorithm is used to segment and cluster the users for the same purpose because of the scattered nature of the data and to find hidden patterns in the data set. To define the number of clusters, the elbow method is used, and customers are grouped with respect to different attributes. Based on the analysis, a decision is made about which set of customers to target. © 2023 IEEE.

关键词： K-means clustering

来源：评论

学校读者我要写书评

暂无评论

Analysis of Security Mechanisms Used in ETL Tools 32

Analysis of Security Mechanisms Used in ETL Tools

引用

32nd Telecommunications Forum, TELFOR 2024

作者： Jeftovic, Milica Vuletic, Pavle Stanisavljevic, Zarko Capgemini Engineering Belgrade Serbia University of Belgrade School of Electrical Engineering Department of Computer Science and Information Technology Belgrade Serbia

ISBN: (纸本)9798350391053

This paper analyses current best practices for using security mechanisms in extract, transform, and load (ETL) tools. The results of the analysis are visually represented in a form of an optimal security model used in these tools at the moment. In order to test the recommended model, it was implemented in a selected open source ETL tool. At the end, advantages and disadvantages of the recommended model are enlisted and directions for further improvement of the model are given. © 2024 IEEE.

关键词： ETL tools Security Mechanisms Security Model

来源：评论

学校读者我要写书评

暂无评论

Combining Event Segment Classification and Graph Self-Encoder for Event Prediction 2

Combining Event Segment Classification and Graph Self-Encode...

引用

2nd International Conference on Big Data, Information and computer Network, BDICN 2023

作者： Yuan, Yewen Wang, Yongli Gong, Xiaoze Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing China Unit Baicheng63850 China

ISBN: (纸本)9798350321364

Scripted event prediction is the automatic learning of knowledge from unstructured text to predict subsequent events based on a given sequence of events. Existing work has mainly been based on event pairs or entire event chains for learning, which may ignore much structural and semantic information. In order to learn global dependencies between nodes in event evolution graphs, this paper proposes an event prediction method combining event segment classification and graph self-encoder, which uses a variational self-encoder of the graph to obtain the global structure of the event graph, learn and represent potential evolution patterns between events, and construct a global representation of the event graph. The method combines different rules to classify event segments, learns the features of different event segments, and predicts events by determining the scores of candidate events based on the correlation between the event segments and the candidate events. Experiments on a standard dataset demonstrate the effectiveness of the method. © 2023 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

MRI-Based Biomarker in the Diagnosis of Alzheimer’s Disease Using Attention-UNet

引用

SN computer science 2025年第3期6卷 1-15页

作者： Arumugam, J. Prasanna Venkatesan, V. Beigh, Tabiya Department of Computer Science School of Engineering and Technology Pondicherry University Kalapet Puducherry Puducherry 605014 India Department of Banking Technology School of Management Pondicherry University Kalapet Puducherry Puducherry 605014 India

Dementia can occur in various forms;some types of dementia are curable and some are not. Among the non-curable forms, Alzheimer’s disease (AD) is the prominent one. There is no effective treatment to cure it but it may reduce its impact if diagnosed early. Accurate classification of AD is necessary for diagnosing and giving the right effective treatment to patients. Specific regions in the brain serve as hotspots for AD, potential imaging biomarkers that contribute effectively thereby improving classification accuracy. We introduce a new Attention-based ’U’ shaped Convolution Neural Network in this study to identify imaging biomarkers of AD using 3D T1-weighted MRI data, which excels at distinguishing between gray matter, white matter, and cerebrospinal fluid (CSF). Our model is improved UNet with an enhanced convolution block attention module named EnCBAMUNet. This makes it ideal for brain imaging, where the differentiation between these tissues is crucial for identifying pathology, such as in neurodegenerative diseases. Our method is tested on both ADNI and OASIS datasets. Classification accuracy of 99.8% for Healthy as control normal(CN) vs Alzheimer’s disease(AD) is obtained in binary classification and 95.5% for multi-classifier Alzheimer’s Disease, Mild Cognitive Impairment(MCI) is a transient state from healthy to disease state and Control Normal as Healthy(AD vs MCI vs CN) in multiclass classification. We visualize regions (hippocampus, ventricles, and some parts of cortex) of the brain responsible for AD using the Three Dimensional Gradient-weighted Class Activation Mapping (3D-Grad-CAM) method of our model due to the deep learning model(DL) that is black-box in nature. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Alzhimer disease Classification Deep learning MRI imaging biomarker Visualization

来源：评论

学校读者我要写书评

暂无评论

A Systematic Review of IoT Malware Detection using Machine Learning 10

A Systematic Review of IoT Malware Detection using Machine L...

引用

17th INDIACom;2023 10th International Conference on Computing for Sustainable Global Development, INDIACom 2023

作者： Sharma, Badal Kumar, Rajendra Kumar, Anuj Chhabra, Megha Chaturvedi, Saumya School of Engineering and Technology Sharda University Department of Computer Science and Engineering Greater Noida India

ISBN: (纸本)9789380544472

The Internet of Things (IoT) is spreading quickly across the globe, yet its security is insufficient. The results are encouraging for using machine learning to identify malware on these frequently unattended devices. Among all infections found in mobile networks, the Internet of Things was the target of more than 30% of them. Such detection in a single IoT node has been done yet, though, because IoT nodes often have insufficient processing. This study explores the feasibility of malware detection in a single IoT device by scaling machine learning methods to allow a single IoT device to perform near-real- time network traffic anomaly detection and label packets as malware or benign. Using one of the machine learning techniques, the suggested software for categorizing data points from the IoT-23 dataset can be implemented on an ESP32 device. When completely deployed, this could imply that IoT devices would be able to decide for themselves whether a network connection is a part of a malware attack or a regular connection. Interconnected devices and terminals are used as a lucrative target by cybercriminals, especially those who have the backing of the government or other national state actors. Vector malware application is a frequent form of assault. Using the operational code sequence or functional code of the device, the Learning Things and Learning technique seeks out harmful online applications on the Internet. This paper uses deep learning approach to convert the Opcodes into vector space and categorize them into dangerous and malicious Applications. This paper present evidence for the efficiency of our proposed approach against malware detection and garbage Code insertion assaults. © 2023 Bharati Vidyapeeth, New Delhi.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Fake News Detection Using Boosting Ensemble Method 14

Fake News Detection Using Boosting Ensemble Method

引用

14th International Conference on Cloud Computing, Data science and engineering, Confluence 2024

作者： Kaur, Kirandeep Kashyap, Nirbhay Amity School of Engineering and Technology Amity University Computer Science Department Uttar Pradesh Noida201301 India

ISBN: (纸本)9798350344837

Around the globe, Fake news is a raising concern and has several negative repercussions on people, societies, and countries. It has several detrimental effects on society, including disseminating false information and causing confusion, inciting social unrest and division, eroding credibility and confidence, endangering public health, and causing political instability. To protect people from damage, preserve informational accuracy, stop the spread of hate speech, support a vibrant democracy, and maintain journalism's reputation, fake news detection is crucial. In this work, we attempt to tackle this problem by putting forth a method that uses machine learning classification to distinguish between authentic and fraudulent content. Natural language processing and machine learning ensemble approach are taken into consideration for this. In natural language processing, TF-IDF and Word2vec features are used and optimized using wrapper method (via RFE algorithm). Training and testing are done on 'Fake News Detection' dataset from Kaggle. Exploratory examination upon the boosting ensemble method tells that our learners perform better than other options. © 2024 IEEE.

关键词： Statistical tests

来源：评论

学校读者我要写书评

暂无评论

Combining Feature Fusion Link Prediction with Graph Neural Networks 2

Combining Feature Fusion Link Prediction with Graph Neural N...

引用

2022 2nd Conference on High Performance Computing and Communication engineering, HPCCE 2022

作者： Zheng, Yamin Wang, Yongli Gong, Xiaoze School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing China Unit Baicheng63850 China

ISBN: (纸本)9781510663374

At present, Graph Neural Network (GNN) methods usually follow the node centered message passing process, and rely heavily on smooth node characteristics rather than graph structure. In view of this limitation, based on the heuristic method and graph attention mechanism, a feature fusion link prediction model (SAFL) combined with graph neural network is proposed. This model extracts the enclosing subgraphs around the target, combines the attention mechanism to assign neighbor weights to learn useful structural features, considers the impact of different nodes on the link, and fuses the graph neural network with the characteristics of input nodes to predict the link. The experiment on OGB dataset shows that the proposed link prediction model based on heuristic method enhances the graph structure characteristics, effectively represents the connectivity of enclosing subgraphs, and achieves better performance in link prediction. © 2023 SPIE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

A secure quantum homomorphic encryption ciphertext retrieval scheme

引用

Soft Computing 2025年第3期29卷 1497-1509页

作者： Cheng, Zhen-Wen Chen, Xiu-Bo Xu, Gang Chang, Yan Miao, Li-Hua Yang, Yi-Xian Wang, Ya-Lan School of Data and Computer Science Shandong Women’s University Jinan250300 China Information Security Center State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing100876 China School of Information Science and Technology North China University of Technology Beijing100144 China School of Cybersecurity Chengdu University of Information Technology Chengdu610225 China Huawei Technologies Co. Ltd Shenzhen518129 China Department of Computer Science Faculty of Engineering and Physical Sciences University of Surrey Surrey GuildfordGU2 7XH United Kingdom

In a recent paper (Gong et al. Quantum Inf Process 19:3, 2020), a novel ciphertext retrieval scheme based on the Grover algorithm and quantum homomorphic encryption was presented. In this scheme, when the server performs the operation of marking the solution on the user’s encrypted state in the Grover iteration, it needs to remove many gate-errors generated in the homomorphic evaluation of the T gate. And the server could judge this specific solution from the quantum circuit of marking the solution. It makes this scheme unable to achieve the low-cost and secure ciphertext retrieval. Therefore, we improve the Gong et al.’s scheme and propose a secure quantum homomorphic encryption ciphertext retrieval scheme. In our scheme, the trusted third party is introduced to cooperate with the server to execute the Grover algorithm. In each Grover iteration, the trusted third party can quickly mark the solution on the plaintext state, encrypt the marked state, and transmit it to the server. Then the server performs the remaining operations of this Grover iteration on the encrypted state. The trusted third party finally decrypts the iterated state. This cooperative approach ensures that the number of auxiliary qubits required and extra quantum gates executed in our scheme are lower than the Gong et al.’s scheme. By analyzing the security of our scheme, we confirm that the server and the trusted third party will not be informed of this solution. Thus, our scheme realizes the secure ciphertext retrieval with low computational overhead. We utilize IBM’s Qiskit framework to simulate our scheme, and the experimental result shows that our scheme is correct. It is worth noting that the low-cost and secure ciphertext retrieval will play a crucial role in modern information security and privacy protection. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Ciphertext

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：