检索结果-内蒙古大学图书馆

A three-branch deep neural network for diagnosing respiratory sounds

Neural Computing and Applications 2024年第35期36卷 22611-22631页

作者： Imani, Maryam Ghassemian, Hassan Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran

Finding an accurate model is essential for classification of respiratory pathologies through extraction and fusion of respiratory sounds’ features. To handle the unlabeled data, a sequence of autoencoders are used for data augmentation. A deep neural network framework is proposed to extract three types of features: 1-Features based on the human auditory system, using Mel-frequency cepstral coefficients (MFCCs), 2-Temporal features contained in the sequence of sound signal using the long short-term memory (LSTM) network, and 3-The nonlinear and complex relationship among temporal characteristics in neighborhood regions using a two-dimensional (2D) convolutional neural network (CNN) applied to the sound sequence converted to a 2D time array. Three branches are fused through some fully connected layers. The ICBHI 2017 sound database is used in two cases of 6 and 3 pathological classes. In the case of 6-classes database, 98.72% overall accuracy, 98.46% kappa coefficient, 99.66% sensitivity, 98.70% specificity and 99.70% F1-Sscore is provided. In the case of 3-classes database, 97.18% overall accuracy, 95.77% kappa coefficient, 98.31% sensitivity, 99.16% specificity and 98.93% F1-score are achieved. In contrast to the many researchers that are applying CNN on the Mel spectrogram of the audios, CNN is applied to the 2D form of the time series to find the hidden relationship among the time instants in local regions. Moreover, fusion of these hidden features with the LSTM and MFCC features leads to an accurate multiple classification. This work proposes a new deep learning framework for fusion of three types of sound features, which leads to a significant improvement in multiple respiratory infection diagnosing. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Unsupervised Hyperspectral image Classification: Spatial and Spectral Feature Fusion with Masked Autoencoders 20

Unsupervised Hyperspectral Image Classification: Spatial and...

引用

20th CSI International Symposium on Artificial Intelligence and Signal processing, AISP 2024

作者： Mirsharji, Seyed Ali Ghassemian, Hassan Tarbiat Modares University Image Processing and Information Analysis Lab. Faculty of Electrical and Computer Engineering Tehran Iran

ISBN: (纸本)9798350383942

In this study, we present an innovative unsupervised hyperspectral image classification method using a dual-branch architecture that merges spatial and spectral feature extraction. Our unique approach employs masked autoencoders, significantly outperforming traditional methods with an impressive overall accuracy of 97.1%. The paper details the model's performance evaluation, offers visual insights into its classification capabilities, and compares it with existing techniques, demonstrating its effectiveness and potential for advancing remote sensing applications. © 2024 IEEE.

关键词： image classification

来源：评论

学校读者我要写书评

暂无评论

Unsupervised Change Detection in SAR images Using a Six-Branch CNN and Adaptive Window Approach 31

Unsupervised Change Detection in SAR Images Using a Six-Bran...

引用

31st International Conference on electrical engineering, ICEE 2023

作者： Kakoolvand, Abbas Imani, Maryam Ghassemian, Hassan Tarbiat Modares University Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tehran Iran

ISBN: (纸本)9798350312560

Change detection is one of the important and hot topics in remote sensing. Adaptive windowing approaches can preserve image details while reduce noise in the process of change detection. In the proposed method, two difference images (DIs) are obtained by using the adaptive window approach. Some fake labels are provided from these DIs. Neural networks have a good performance in image processing. To use this advantage, a six-branch convolutional neural network (CNN) is trained using the fake labels. Using these six branches with utilizing the adaptive windows, the network can preserve image detail while reduce noise. The proposed method based on three criteria (PCC, Kappa, and F1 score) has the best results in two datasets and has results close to the best method in other datasets. © 2023 IEEE.

关键词： Synthetic aperture radar

来源：评论

学校读者我要写书评

暂无评论

Fusion of Multi-Level CNN With LBP Features For Facial Emotion Recognition 31

Fusion of Multi-Level CNN With LBP Features For Facial Emoti...

引用

31st International Conference on electrical engineering, ICEE 2023

作者： Bahmanabady, Ehsan Imani, Maryam Ghassemian, Hassan Tarbiat Modares University Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tehran Iran

ISBN: (纸本)9798350312560

A facial emotion recognition framework is proposed in this work. The convolutional neural network (CNN) has high ability in extraction of hierarchical spatial features from low level texture characteristics to high level contextual features. A simple CNN model with three layers is suggested in this paper where the features extracted in all layers containing multi-level features are activated and used for classification. In addition, the local binary pattern (LBP) descriptor is used to extract discriminative features from the spatial structure of the input image. Therefore, four feature sources are provided by multi levels of CNN and the LBP descriptor. Each feature source is used for facial emotion recognition by applying to the support vector machine (SVM) classier. Finally, the majority voting rule is used for decision fusion to provide the final emotional label of each given face image. The proposed method with 84% overall accuracy, 83% weighted F1-score and 81% kappa coefficient provides the best performance compared to LBP, multi-level CNN and two-dimensional principal component analysis (2DPCA) methods. © 2023 IEEE.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

Deep Curriculum Learning for PolSAR image Classification 12

Deep Curriculum Learning for PolSAR Image Classification

引用

12th Iranian/2nd International Conference on Machine Vision and image processing, MVIP 2022

作者： Mousavi, Hamidreza Imani, Maryam Ghassemian, Hassan Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tehran Iran

ISBN: (纸本)9781665412162

Following the great success of curriculum learning in the area of machine learning, a novel deep curriculum learning method proposed in this paper, entitled DCL, particularly for the classification of fully polarimetric synthetic aperture radar (PolSAR) data. This method utilizes the entropy-alpha target decomposition method to estimate the degree of complexity of each PolSAR image patch before applying it to the convolutional neural network (CNN). Also, an accumulative mini-batch pacing function is used to introduce more difficult patches to CNN. Experiments on the widely used data set of AIRSAR Flevoland reveal that the proposed curriculum learning method can not only increase classification accuracy but also lead to faster training convergence. © 2022 IEEE.

关键词： Synthetic aperture radar

来源：评论

学校读者我要写书评

暂无评论

Spatial Quality Assessment of Pansharpened images Based on Gray Level Co-Occurrence Matrix 12

Spatial Quality Assessment of Pansharpened Images Based on G...

引用

12th Iranian/2nd International Conference on Machine Vision and image processing, MVIP 2022

作者： Aghapour Maleki, Shiva Ghassemian, Hassan Tarbiat Modares University Image Processing and Information Analysis Laboratory Faculty of Electrical and Computer Engineering Tehran Iran

ISBN: (纸本)9781665412162

Assessing the quality of pansharpened images is a critical issue in order to obtain a quantitative score to represent the quality and compare the performance of different fusion methods. Most of the introduced metrics for pansharpened image quality assessment, evaluate the spectral content of the image, while in different applications of remote sensing like detection and identification of image objects, spatial quality has an important role. In the current study, a new index for spatial quality assessment is introduced that extracts gray level co-occurrence matrix (GLCM) from distorted and reference images and compares the similarities of these features. The tempere image database 2013 (TID2013) that provides reference and different types of distorted images with subjective scores of each image is used as the desired database. To solve the high computational complexity of obtaining GLCM features, the fast GLCM method is employed. In this way, 16 different features are extracted. To select the features that have the most consistency with the human visual system (HVS), the forward floating search method is used as a feature selection method and five features are obtained as the final features to form the desired index. Experimental results show the efficiency of the proposed method in determining the spatial quality of fused images compared with that of the available quality assessment metrics. © 2022 IEEE.

关键词： Remote sensing

来源：评论

学校读者我要写书评

暂无评论

Ovarian Tumor Ultrasound image Segmentation with Deep Neural Networks

Ovarian Tumor Ultrasound Image Segmentation with Deep Neural...

引用

Iranian Machine Vision and image processing (MVIP)

作者： Mahnaz Siahpoosh Maryam Imani Hassan Ghassemian Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran

The precise and automated segmentation of ovarian tumors in medical images plays a pivotal role in the treatment of ovarian cancer in women. U-Net has demonstrated remarkable success in the field of medical image segmentation. However, due to its small receptive field, U-Net faces challenges in extracting global context information. Moreover, due to the significant variation in scale and size among tumors, it is essential to employ a network capable of effectively extracting information at Multiple scales. In this study, we present a U-Net-based network named PCU-Net for the segmentation of ovarian tumors, incorporating ConvMixer and Pyramid Dilated Convolution (PDC) modules. The ConvMixer module captures global context information by utilizing large-size kernels. The PDC module integrates local and global contextual patterns through utilization of parallel dilated convolution with different dilation rate. Furthermore, our model has fewer parameters than U-Net. We assess the proposed method’s performance using the Multi-Modality Ovarian Tumor Ultrasound (MMOTU) dataset. The results indicate that in comparison to U-Net, our proposed PCU-Net exhibits an improvement of 4.23% in terms of Intersection over Union (IoU) and 2.99% in terms of Dice Similarity Coefficient (DSC).

关键词：

来源：评论

学校读者我要写书评

暂无评论

Weighted Features Based Classification of Polarimetric SAR images

Weighted Features Based Classification of Polarimetric SAR I...

引用

Iranian Machine Vision and image processing (MVIP)

作者： Fatemeh Saneipour Maryam Imani Hassan Ghassemian Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran

Today, classification of polarimetric images is an important topic where various statistical pattern recognition methods have been used to achieve the high accurate classification maps. In this work, weighting the polarimetric features according to their statistical behavior (the mean vector and variance values as the first and second statistics) is suggested to improve the PolSAR image classification. A weighted feature matrix is composed and applied to the popular classifiers such as maximum likelihood, K-nearest neighbor and support vector machine. The weighted feature matrix can be also implemented on other arbitrary classifiers to improve their discrimination ability. The experiments on the L-band AIRSAR dataset show appropriate classification results.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Spatial-Spectral Neural Network for High Resolution Multispectral image Classification

Spatial-Spectral Neural Network for High Resolution Multispe...

引用

Iranian Machine Vision and image processing (MVIP)

作者： Roozbeh Tanha Hassan Ghassemian Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran

Classification of multispectral images in remote-sensing area having the capability to analyze and categorize diversified land cover. In this issue, extracting suitable spatial, spectral and even temporal features is one of the main challenges. Also, the existence of sufficient data required for the classification training process is another challenge, because in many cases it may not be available and we may not even have a reliable classification map. The use of neural networks for simultaneous feature extraction and classification is very popular and significant progress has been made in this field, but these networks usually have a high computational cost and require significant training data in the training process. In this work we propose a neural network for multispectral image classification purpose which requires few training samples and less calculation without using filterbanks for spatial feature extraction and it can improve classification accuracy by fusion of spatial and spectral features. The simulations indicate that the proposed method shows an acceptable performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Spectral-Spatial Anomaly Detection in Hyperspectral imagery Based on Dual-Domain Autoencoders

Spectral-Spatial Anomaly Detection in Hyperspectral Imagery ...

引用

Iranian Machine Vision and image processing (MVIP)

作者： Mohamad Ebrahim Aghili Hassan Ghassemian Maryam Imani Arani Image Processing and Information Analysis Lab Faculty of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran

Hyperspectral anomaly detection is crucial for applications like aerial surveillance in remote sensing images. However, robust identification of anomalous pixels remains challenging. A novel spectral-spatial anomaly detection technique called Dual-Domain Autoencoders (DDA) is proposed to address these challenges. First, Nonnegative Matrix Factorization (NMF) is applied to decompose the hyperspectral data into anomaly and background components. Refinement of the designation is then done using intersection masking. Next, a spectral autoencoder is trained on identified background signature pixels and used to reconstruct the image. The reconstruction error highlights spectral anomalies. Furthermore, a spatial autoencoder is trained on principal component patches from likely background areas. Fused reconstruction error from the spectral and spatial autoencoders is finally used to give enhanced anomaly detection. Experiments demonstrate higher AUC for DDA over individual autoencoders and benchmark methods. The integration of matrix factorization and dual-domain, fused autoencoders thus provides superior anomaly identification. Spatial modeling further constrains the background, enabling accurate flagging of unusual local hyperspectral patterns. This study provides the effectiveness of employing autoencoders trained on intelligently sampled hyperspectral pixel signatures and spatial features for improved spectral-spatial anomaly detection.

关键词：

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：