检索结果-内蒙古大学图书馆

25th IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP 2000

作者： Necioǧlu, Burhan F. Clements, Mark A. Barnwell, Thomas P. Center for Signal and Image Processing School of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA30332 United States

ISBN: (纸本)0780362934

This paper describes a method for the unsupervised and gender-independent estimation of the average human vocal tract length from the speech waveform, and reports results obtained on Fant's (1960) X-ray vowel data as well as results from experiments performed on multiple sentence utterances of 86 male and 78 female TIMIT speakers, including correlation analyses between the vocal tract length estimates and given body heights. The investigated error criteria that make non-iterative, closed-form estimator solutions possible are all found to achieve good speaker clustering potential for both male and female subgroups. © 2000 IEEE.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Self-supervised model-informed deep learning for low-SNR SS-OCT domain transformation

引用

Scientific Reports 2025年第1期15卷 1-20页

作者： Sajed Rakhshani Mahnoosh Tajmirriahi Farnaz Sedighin Hossein Rabbani Amirali Arbab Aref Habibi Mohsen Pourazizi Medical Image and Signal Processing Research Center School of Advanced Technologies in Medicine Isfahan University of Medical Sciences Isfahan Iran Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan Iran Department of Ophthalmology Isfahan Eye Research Center Isfahan University of Medical Sciences Isfahan Iran

This article introduces a novel deep-learning based framework, Super-resolution/Denoising network (SDNet), for simultaneous denoising and super-resolution of swept-source optical coherence tomography (SS-OCT) images. The novelty of this work lies in the hybrid integration of data-driven deep-learning with a model-informed noise representation, specifically designed to address the very low signal-to-noise ratio (SNR) and low-resolution challenges in SS-OCT imaging. SDNet introduces a two-step training process, leveraging noise-free OCT references to simulate low-SNR conditions. In the first step, the network learns to enhance noisy images by combining denoising and super-resolution within noise-corrupted reference domain. To refine its performance, the second step incorporates Principle Component Analysis (PCA) as self-supervised denoising strategy, eliminating the need for ground-truth noisy image data. This unique approach enhances SDNet’s adaptability and clinical relevance. A key advantage of SDNet is its ability to balance contrast-texture by adjusting the weights of the two training steps, offering clinicians flexibility for specific diagnostic needs. Experimental results across diverse datasets demonstrate that SDNet surpasses traditional model-based and data-driven methods in computational efficiency, noise reduction, and structural fidelity. The framework excels in improving both image quality and diagnostic accuracy. Additionally, SDNet shows promising adaptability for analyzing low-resolution, low-SNR OCT images, such as those from patients with diabetic macular edema (DME). This study establishes SDNet as a robust, efficient, and clinically adaptable solution for OCT image enhancement addressing critical limitations in contemporary imaging workflows.

关键词： Denoising Low-SNR OCT Model-aware deep learning Self-supervised learning Super-resolution

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-Enabled, Computed Tomography-Based Race- and Sex-Specific Epicardial Adipose Tissue Thresholds for Cardiovascular Risk Stratification

Deep Learning-Enabled, Computed Tomography-Based Race- and S...

引用

Medical Imaging 2025: Clinical and Biomedical Imaging

作者： Buchwald, Mikolaj Shanbhag, Aakash Miller, Robert J.H. Zhang, Wenhao Michalowska, Anna M. Killekar, Aditya Lemley, Mark Kavanagh, Paul B. Fujito, Hidesato Liang, Joanna X. Builoff, Valerie Knight, Stacey Einstein, Andrew J. Miller, Edward J. Feher, Atilla Sinusas, Albert J. Chareonthaitawee, Panithaya Bullock-Palmer, Renee P. Di Carli, Marcelo F. Berman, Daniel S. Dey, Damini Slomka, Piotr J. Departments of Medicine Division of Artificial Intelligence in MedicineImaging and Biomedical Sciences Cedars-Sinai Medical Center Los AngelesCA United States Poznan Supercomputing and Networking Center Polish Academy of Sciences Poznan Poland Signal and Image Processing Institute Ming Hsieh Department of Electrical and Computer Engineering University of Southern California Los AngelesCA United States Department of Cardiac Sciences University of Calgary CalgaryAB Canada Center of Radiological Diagnostics National Medical Institute The Ministry of the Interior and Administration Warsaw Poland Department of Cardiology Nihon University Itabashi Hospital Tokyo Japan Intermountain Medical Center Heart Institute MurrayUT United States Division of Cardiology Columbia University Irving Medical Center New York CityNY United States Section of Cardiovascular Medicine Department of Internal Medicine Yale University School of Medicine New HavenCT United States Department of Cardiovascular Medicine Mayo Clinic RochesterMN United States Department of Cardiology Deborah Heart and Lung Center Browns MillsNJ United States Division of Cardiovascular Medicine Department of Medicine Brigham and Women’s Hospital Harvard Medical School BostonMA United States Cardiovascular Imaging Program Departments of Radiology and Medicine Division of Nuclear Medicine and Molecular Imaging Department of Radiology Brigham and Women’s Hospital Harvard Medical School BostonMA United States

ISBN: (纸本)9781510685987

While the volume of epicardial adipose tissue (EAT) has been linked to various conditions and showed a prognostic value of cardiovascular events, an insufficient effort was put into proposing clear, race- and sex-specific thresholds of high-risk and low-risk EAT amounts. Also, previous works focused on the absolute volume of EAT in milliliters, adjusted to the body size of the patients, and none has proposed a relative measure of what is the amount of EAT for a given patient also in reference to the size of the patient’s heart. In this retrospective study, computed tomography attenuation correction (CTAC) scans from 20,587 patients (58% male;84% White, 13% Black, 3% Asian) from 6 sites were used. With the previously described deep learning (DL)-based method, EAT values per patient were calculated, and subsequently normalized using either the DL-obtained cardiac volume mask, body size, or both. The association between the percent of EAT in reference to the heart size and incident mortality or nonfatal myocardial infarction (MI) was evaluated with Cox models adjusted for age, sex, body mass index, race, cigarette smoking status, hypertension, diabetes, family history of coronary artery disease, dyslipidemia, and left ventricle size. In total, 4,133 events were observed over a period of 6-year follow-up. Our analysis showed that when the amount of EAT was normalized jointly for body surface area and heart size, the predictive effect of EAT was diminished (p=0.85). Patients with higher EAT density, however, were more likely to experience death or MI during follow-up (p © 2025 SPIE.

关键词： Risk assessment

来源：评论

学校读者我要写书评

暂无评论

Temporal partition particle filtering for multiuser detectors with mutually orthogonal sequences

Temporal partition particle filtering for multiuser detector...

引用

2005 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP '05

作者： Özgür, Soner Williams, Douglas B. Center for Signal and Image Processing School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta GA 30332-0250 United States

ISBN: (纸本)0780388747

We propose a blind multiuser detector based on Monte Carlo Markov chain (MCMC) techniques. The detector exploits mutually orthogonal complementary sequences to distinguish between transmitting users and space-time codes to take advantage of the available spatial diversity. We propose a partitioning scheme for the symbol draws in the MCMC algorithm that reduces the complexity without any degradation in performance. The detector's performance is simulated in an iterative receiver that utilizes an outer coder. The simulations display some loss in coding gain because of the blind nature of the system;however, diversity gain is preserved. © 2005 IEEE.

关键词： Detectors

来源：评论

学校读者我要写书评

暂无评论

MIMO transmission subspace tracking with low rate feedback

MIMO transmission subspace tracking with low rate feedback

引用

2005 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP '05

作者： Yang, Jingnong Williams, Douglas B. Center for Signal and Image Processing School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta GA 30332-0250 United States

ISBN: (纸本)0780388747

This paper describes a low-rate feedback algorithm for conveying partial channel state information - specifically, the dominant row subspace of the channel matrix - from the receiver to the transmitter in a continuously time-varying multiple-antenna environment. Since subspaces are points on a complex Grassmann manifold, variations in subspaces are treated as a piecewise geodesic process on the manifold. The receiver feeds back one bit to indicate the preferred sign of a random velocity matrix of the geodesic. Numerical results show that the performance of the proposed algorithm is better than the Grassmannian subspace packing approach at low-to-medium Doppler frequency and always better than the previously proposed gradient sign feedback scheme. © 2005 IEEE.

关键词： Communication channels (information theory)

来源：评论

学校读者我要写书评

暂无评论

A ROBUST APPROACH FOR HIGH-RESOLUTION FREQUENCY ESTIMATION

引用

IEEE TRANSACTIONS ON signal processing 1991年第3期39卷 627-643页

作者： SANG, GO KASHYAP, RL School of Electrical Engineering Purdue University West Lafayette IN USA Signal Processing Branch Computer Science Center Texas Instruments Inc. Dallas TX USA

We investigate a robust estimation method for estimating frequencies of received signals. The received signals can be represented as a sum of sinusoidal signals and an additive noise process. The additive noise is assumed to be a mixture of a Gaussian and an outlier process. It is known that a robust estimate performs much better than conventional estimates such as least square estimates (LSE) or Gaussian maximum likelihood estimates (MLE-G), when the noise has a mixture distribution or when the observations have a low signal-to-noise ratio (SNR). Even a single outlier can cause large errors in conventional estimation methods because they are sensitive to minor deviations from the underlying assumptions. Among the family of robust estimates, we consider the M-estimate which is obtained by the minimization of a nonquadratic function of normalized residuals. The following are the topics considered in our paper. First, the influence function of a robust estimate is derived for time varying (sinusoidal) signals. Second, the variance of the robust estimate is derived through the influence function. It is found that the robust estimate attains the Cramer-Rao lower bound (CRLB) for the contaminated Gaussian distribution. It is of order O(N-3) and is close to the CRLB for the perfect Gaussian distribution. Finally, we introduce some basic definitions for the high resolution frequency estimation method and prove that the robust estimate has the high resolution property. This property is not possessed by estimation methods, such as autoregressive based methods.

关键词： Frequency estimation Maximum likelihood estimation Noise robustness Gaussian noise signal to noise ratio signal processing Additive noise Least squares approximation Gaussian distribution signal resolution

来源：评论

学校读者我要写书评

暂无评论

Ultra low bit rate speech coding using an ergodic hidden Markov model

Ultra low bit rate speech coding using an ergodic hidden Mar...

引用

2005 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP '05

作者： Lee, Matthew E. Durey, Adriane Swalm Moore, Elliot Clements, Mark Georgia Institute of Technology Center for Signal and Image Processing School of Electrical and Computer Engineering Atlanta GA 30332-0250 United States

ISBN: (纸本)0780388747

This paper presents the framework for an ultra low bit rate speech vocoder. The system is based on a recognition-synthesis paradigm in which a single ergodic hidden Markov model (EHMM) is used to capture the statistical characterizations of speech in a flexible manner capable of limiting the effects of recognition errors. Because predetermined speech units are not used, this system has the advantage of not requiring a transcription for the training data set. By incorporating a mixed excitation scheme based on an improved MELP formulation into the EHMM, additional gains in quality and speaker characterization are achieved at no cost to the bit rate. © 2005 IEEE.

关键词： Speech coding

来源：评论

学校读者我要写书评

暂无评论

Near field imaging of subsurface targets using wide-band multi-static RELAX/CLEAN algorithms

Near field imaging of subsurface targets using wide-band mul...

引用

2005 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP '05

作者： Alam, Mubashir McClellan, James H. Center of Signal and Image Processing School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta GA 30332-0250 United States

ISBN: (纸本)0780388747

This paper presents two new imaging algorithms for detecting the positions of subsurface targets, e.g., land mines, using seismic waves. They are based on the CLEAN algorithm and its high resolution version RELAX. This paper will show that how the CLEAN and RELAX algorithms can be modified to work in a multi-static active array setup for detecting passive targets. Seismic surface waves reflected from various targets are collected at a receiving array and processed to locate the reflectors. The modified imaging algorithms will be demonstrated to work for experimental seismic data that includes mines and rocks (clutter). © 2005 IEEE.

关键词： Imaging techniques

来源：评论

学校读者我要写书评

暂无评论

Synchronous and asynchronous distributed DSP education 14

Synchronous and asynchronous distributed DSP education

引用

14th International Conference on Digital signal processing, DSP 2002

作者： Hayes, Monson H. Jackson, JoeI R. Center for Signal and Image Processing Georgia Institute of Technology School of Electrical and Computer Engineering AtlantaGA30332-0250 United States

ISBN: (纸本)0780375033

The Georgia Tech Regional engineering Program (GTREP) was originally created to provide the opportunity for students in southeastern Georgia to earn a Georgia Tech undergraduate engineering degree without leaving the region. Students complete two years of general education and engineering prerequisites at their home institutions before beginning the Georgia Tech curriculum as juniors. Georgia Tech courses are taught both by local faculty in southeast Georgia and remotely by faculty on the main campus. This system is further complicated by the fact that courses are delivered to three partner institutions as well as the GTREP facilities. The GTREP program is an ideal test bed for novel distributed education methods. In this paper, we present some results for the combined approach we have developed, and outline plans for future scaling as the GTREP program expands. © 2002 IEEE.

关键词： Students

来源：评论

学校读者我要写书评

暂无评论

Hierarchical and Multimodal Data for Daily Activity Understanding

arXiv

引用

arXiv 2025年

作者： Kaviani, Ghazal Yarici, Yavuz Kim, Seulgi Prabhushankar, Mohit AlRegib, Ghassan Solh, Mashhour Patil, Ameya OLIVES Center for Signal and Information Processing CSIP School of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA United States Amazon Lab126 San FranciscoCA United States

Daily Activity Recordings for artificial intelligence (DARai, pronounced /Dahr-ree/), is a multimodal, hierarchically annotated dataset constructed to understand human activities in real-world settings. DARai consists of continuous scripted and unscripted recordings of 50 participants in 10 different environments, totaling over 200 hours of data from 20 sensors including multiple camera views, depth and radar sensors, wearable inertial measurement units (IMUs), electromyography (EMG), insole pressure sensors, biomonitor sensors, and gaze tracker. To capture the complexity in human activities, DARai is annotated at three levels of hierarchy: (i) high-level activities (L1) that are independent tasks, (ii) lower-level actions (L2) that are patterns shared between activities, and (iii) fine-grained procedures (L3) that detail the exact execution steps for actions. The dataset annotations and recordings are designed so that 22.7% of L2 actions are shared between L1 activities and 14.2% of L3 procedures are shared between L2 actions. The overlap and unscripted nature of DARai allows counterfactual activities in the dataset. Experiments with various machine learning models showcase the value of DARai in uncovering important challenges in human-centered applications. Specifically, we conduct unimodal and multimodal sensor fusion experiments for recognition, temporal localization, and future action anticipation across all hierarchical annotation levels. To showcase the shortcomings of individual sensors, we conduct domain-variant experiments that are possible because of DARai’s multi-sensor and counterfactual activity design setup. The code, documentation, and dataset is available at the dedicated DARai website. © 2025, CC BY.

关键词： image segmentation

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：