检索结果-内蒙古大学图书馆

International Conference on Bio-Inspired computing: Theories and Applications, BICTA

作者： Ying Li Qiang Zhang Na Ma Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

This paper presents a stratified image hiding scheme based on chaotic algorithm. In the pretreatment process chaotic sequences are used to encrypt the secret image. The histogram of the processed image is similar to that of white noise. A stratified secret data embedding thought is adopted in the embedding process. As the experimental results show, this algorithm has relatively large embedding capacity, good invisibility and robustness.

关键词： Cryptography Algorithm design and analysis Discrete wavelet transforms PSNR Pixel

来源：评论

学校读者我要写书评

暂无评论

A multilevel image encryption algorithm based on chaos and DNA coding

A multilevel image encryption algorithm based on chaos and D...

引用

International Conference on Bio-Inspired computing: Theories and Applications, BICTA

作者： Qian Wang Qiang Zhang Changjun Zhou Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

Chaos has been attracted much attention in current research of image encryption. In this paper, we change the location of pixels and pixel values through the combination of chaos with DNA coding. First of all, we transform the digital image into DNA coding matrix based on DNA code rules. Secondly, we divided this matrix into four sub matrixes, and disorder these sub matrixes through logistic chaotic sequence. Then, we scramble the sub matrixes by adding new logistic chaotic sequences and get the new sub matrixes. Finally, we join all scrambled DNA sub-matrixes in a DNA sequence and decode the scramble DNA matrix to image matrix. The experiment results demonstrate that the proposed scheme is effective, and it has wide secret key space, strong sensibility, high security and good ability of resisting statistic attack.

关键词： Cryptography Chaos DNA Image coding Sequences Logistics Digital images Decoding Security Statistics

来源：评论

学校读者我要写书评

暂无评论

Semi-Supervised Semantic Segmentation Based on Class Balanced Framework

Semi-Supervised Semantic Segmentation Based on Class Balance...

引用

IEEE International Conference on Signal and Image Processing (ICSIP)

作者： Jiacheng Yang Jing Dong Pengfei Yi Rui Liu Ling Wang The Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

ISBN: (数字)9798350350920

ISBN: (纸本)9798350350937

Semi-supervised semantic segmentation is a task that utilizes a small amount of labeled data and a large amount of unlabeled data to achieve pixel-level annotation. Due to the common long tail distribution phenomenon in semi-supervised semantic segmentation datasets and the fixed threshold strategy commonly used in filtering pseudo labels, the model tends to favor the head classes when generating pseudo labels, resulting in poor segmentation performance on tail data. This article proposes a semi-supervised semantic segmentation method based on a class- balanced framework, which improves the model's recognition ability for minority classes and the quality of pseudo labels through tail classes oversampling and dynamic threshold adjustment. This method first constructs a Class Prediction Sampling-based Self-balancing Module (CPS 2 M). It identifies tail classes through Intersection-over-Union (IoU) analysis and constrains the random oversampling process based on the tail classes data distribution degree. Subsequently, the sampled labeled data is mixed with unlabeled samples to transfer prior knowledge, increase training frequency for tail classes, and solve the experience mismatch problem. Finally, we design a Dynamic Threshold Updating module (DTU) based on a non-linear function. The confidence threshold can dynamically change according to the model's learning ability to avoid discarding correct tail classes with low confidence, which alleviates the class imbalance problem of generating pseudo-labels. Based on the DeepLabV3+ network framework and ResNet-101 pre-training model, this method achieves 82.1 mIoU under 1464 labeled partition protocols on the Classic Pascal VOC 2012 dataset, demonstrating the most advanced performance compared to existing methods.

关键词： Training Protocols Filtering Annotations Semantic segmentation Semantics Tail

来源：评论

学校读者我要写书评

暂无评论

Cross-Stage Mutual Distillation for Replay Attack Detection

Cross-Stage Mutual Distillation for Replay Attack Detection

引用

6th International Conference on intelligent computing and Signal Processing (ICSP)

作者： Yinqing Cheng Lasheng Zhao Ling Wang Han Wang Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

ISBN: (数字)9798350376548

ISBN: (纸本)9798350376555

Automated speaker verification systems face a higher risk of replay attacks in practice. However, the existing studies face the problem of limited detection capabilities and insufficient use of shallow fine-grained information. To address these issues, we propose the cross-stage mutual distillation(CS-MD) framework, which involves two models learning from a deep network output of each other in different stages of training. This mutual learning approach enhances the ability of shallow networks to capture fine-grained speech information. Additionally, we use an attentional feature fusion module to integrate shallow information more effectively. The multi-scale attention mechanisms in this module can combine local and global speech features while preserving detailed information. Experimental results on the ASVspoof 2019 physical access dataset demonstrate that our proposed method outperforms state-of-the-art methods in terms of EER and min t-DCF metrics, validating the effectiveness of our CS-MD framework.

关键词： Training Voice activity detection Measurement Knowledge engineering Attention mechanisms Speech enhancement Signal processing Feature extraction Faces

来源：评论

学校读者我要写书评

暂无评论

Image encryption algorithm based on DNA biological properties and chaotic systems

Image encryption algorithm based on DNA biological propertie...

引用

International Conference on Bio-Inspired computing: Theories and Applications, BICTA

作者： Qian Wang Qiang Zhang Xiaopeng Wei Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian university Dalian China

With the rapid development of DNA computing, people began to use DNA technology to encrypt information. In this paper, we combine the biological characteristics of DNA and the improved sequence of Logistic mapping method, encrypt the image. The biological characteristics of DNA can better scramble image pixel, and the improved sequence can better change the pixel values. The encryption effect of this paper is better than traditional encryption, and the correlation of adjacent pixels is very low.

关键词： Biomedical imaging Cryptography Clocks Biology

来源：评论

学校读者我要写书评

暂无评论

Robust keyword Spotting Based on Long-Term Feature Fusion and Band-Weighted Normalization

Robust Keyword Spotting Based on Long-Term Feature Fusion an...

引用

Computer Information Science and Application Technology (CISAT), International Conference on

作者： Zepeng Qiu Lasheng Zhao Ling Wang Tingting Zhang Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

ISBN: (数字)9798350375107

ISBN: (纸本)9798350375114

Currently, convolution-based models for keyword spotting focus predominantly on research conducted in clean speech environments. However, their recognition accuracy is lower in low signal-to-noise ratio conditions. To enhance model robustness, we propose the long-term feature fusion module. By extracting and fusing long-term features in both the time and frequency domains, the model’s perception of long-term features is strengthened, improving its recognition performance in noisy environments. At the same time, we propose a band-weighted normalization method, allowing the model to adjust the importance of different frequency bands during the normalization process, further enhancing the robustness of the model. Experimental results indicate that, on the Google Speech Command V2 dataset, the proposed model achieves higher recognition accuracy under various signal-to-noise ratios compared to the comparative models, with a lower parameter count. Moreover, the proposed model exhibits superior generalization to signal-to-noise ratios not covered during the training phase.

关键词： Training Time-frequency analysis Accuracy Convolution Speech recognition Feature extraction Robustness Internet Noise measurement Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Acoustic Word Embedding Model with Transformer Encoder and Multivariate Joint Loss

Acoustic Word Embedding Model with Transformer Encoder and M...

引用

intelligent computing and Wireless Optical Communications (ICWOC), International Conference on

作者： Yunyun Gao Qiang Zhang Lasheng Zhao Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

As a representation of speech information, acoustic word embedding can enable query-by-example keyword search with low-resource speech data. An acoustic word embedding model with Transformer encoder and multivariate joint loss is presented in this article. First of all, on the basis of BLSTM, the model can extract richer information by adding a Transformer encoder, which improves representation ability of the embedding vector. Then, the contrast loss is improved according to the multiple negative samples, and combined with the hinge loss with distance characteristics, the model generates a more compact vector representation and improves the identification of features. Finally, considering the unbalanced characteristics of samples, the anti-focal loss is introduced to form a multivariate joint loss together with the above two losses, so as to make the model performance better. The model has achieved certain results in word recognition tasks and has certain competitiveness compared with other acoustic word models.

关键词：

来源：评论

学校读者我要写书评

暂无评论

An improved collision detection algorithm based on OBB

引用

Computer Modelling and New Technologies 2014年第1期18卷 71-79页

作者： Man, Ranran Zhou, Dongsheng Zhang, Qiang Key Laboratory of Advanced Design and Intelligent Computing Dalian University Ministry of Education Dalian China

In this paper, we present an effective algorithm based on the Oriented Bounding Box (OBB). Particularly, the article focus on how to reduce the amount of the time that the intersection of bounding boxes are detected for increasing the efficiency of collision detection algorithm through the following three steps. Firstly, the detection time of bounding box intersection is decreased by a pre-treatment method. Secondly, we optimize the traversal approach of the bounding box tree and reduce the depth of the two fork tree under take into account the temporal and spatial correlation of motion of virtual environment. Then, we reduce the breadth of tree result from this algorithm compares the distance between the bounding boxes and traverse the node of distance that near the tree. At last, the validity of the algorithm is verified by programming simulations.

关键词： Virtual reality

来源：评论

学校读者我要写书评

暂无评论

On the lower bounds of DNA word sets for DNA computing

On the lower bounds of DNA word sets for DNA computing

引用

International Conference on Bio-Inspired computing: Theories and Applications, BICTA

作者： Bin Wang Qiang Zhang Xiaopeng Wei Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education Dalian University Dalian China

DNA computing is a new research field, which has been paid attention to by more and more experts with the DNA computing bringing the flying development of mathematics, computer science and other subjects. Because the DNA coding could decrease the emergence of false negative and false positive, it affects the reliability and the accuracy of DNA computing. In the DNA coding, DNA word sets could provide the best and effective DNA sequences which are used in DNA computing to improve the reliability and accuracy of biochemical reaction. However, there are no entire and exact values which satisfy combinational constraints, which induce many problems in DNA coding. For this reason, in this paper, we use the dynamic genetic algorithm to improve lower bounds of DNA word sets which satisfy combinational constraints. By comparing our experimental results with previous researchful work, the results improve the lower bounds and further shorten the value range of DNA word sets.

关键词： DNA computing Sequences Biology computing Biological information theory Computer science Mathematics Genetic algorithms design methodology Thermodynamics Computer science education

来源：评论

学校读者我要写书评

暂无评论

A novel robust digital image watermarking using LPM and HVS

A novel robust digital image watermarking using LPM and HVS

引用

The 2013 3rd International Conference on Photonics and Image in Agriculture Engineering (PIAGENG 2013)

作者： Qiang Zhang Mingli Zhang Xiaopeng Wei Key Laboratory of Advanced Design and Intelligent Computing (Dalian University) Ministry of EducationDalian University

ISBN: (纸本)9780819495587

To resist on general attacks, such as compression, noising and geometric distortion attacks is a common requirement for digital image watermarking or data hiding techniques comparing with rotation, scaling and translation(RST) attacks which are considered to be more challenging. Based on invariant centroid and LPM which is also used to analyze the angle of rotation and ratio of scaling influence after attacks of watermarked image combining with chaotic system and a novel adaptive coefficient α, this paper presents a robust and invisible digital image watermarking technology. The experiment results show that this method has high invisibility and robust, especially withstand RST attacks.

关键词： watermarking LPM α mix-chaotic system invariant centroid

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：