2012 IEEE International Workshop on Machine Learning for Signal Processing最新文献_第5页

Efficient high dynamic range imaging via matrix completion 通过矩阵完成高效的高动态范围成像

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349802

Grigorios Tsagkatakis, P. Tsakalides

{"title":"Efficient high dynamic range imaging via matrix completion","authors":"Grigorios Tsagkatakis, P. Tsakalides","doi":"10.1109/MLSP.2012.6349802","DOIUrl":"https://doi.org/10.1109/MLSP.2012.6349802","url":null,"abstract":"Typical digital cameras exhibit a limitation regarding the dynamic range of the scene radiance they can capture. High Dynamic Range (HDR) imaging refers to methods and systems that aim to generate images that exhibit higher dynamic range between the lightest and the darkest parts of the an image. A typical approach for generating HDR images is exposure bracketing where multiple frames, each one with a different exposure setting, are captured and combined to a HDR image of the scene. The large number of images that exposure bracketing requires often leads to motion artefacts that limit the visual quality of the resulting HDR image. In this work, we propose a novel approach in HDR imaging that significantly reduces the necessary number of images. In our proposed system, we employ the notion of random exposure where each pixel of a single frame collects light for a random amount of time. By collecting a small number of such images, the full sequence of low dynamic range images can be reconstructed and subsequently used for HDR generation. The problem is solved by casting the reconstruction of the sequence as a nuclear norm minimization problem following the premises of the recently proposed theory of Matrix Completion. Experimental results suggest that the proposed method is able to reconstruct the sequence from as low as 20% of the images that traditional techniques require with minimal reduction in image quality.","PeriodicalId":262601,"journal":{"name":"2012 IEEE International Workshop on Machine Learning for Signal Processing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129565495","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 7

Stochastic triplet embedding 随机三联体嵌入

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349720

L. Maaten, Kilian Q. Weinberger

引用次数: 205

Optimal cost function and magnitude power for NMF-based speech separation and music interpolation 基于nmf的语音分离和音乐插值的最优代价函数和幅度功率

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349726

Brian King, C. Févotte, P. Smaragdis

引用次数: 45

Markov Chain Monte Carlo inference for probabilistic latent tensor factorization 概率潜张量分解的马尔可夫链蒙特卡罗推理

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349799

Umut Simsekli, A. Cemgil

引用次数: 6

Single-sided objective speech intelligibility assessment based on Sparse signal representation 基于稀疏信号表示的单面客观语音可理解度评估

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349776

G. Costantini, M. Todisco, R. Perfetti, A. Paoloni, G. Saggio

引用次数: 4

Neural correlates of visual perception in rapid serial visual presentation paradigms 快速序列视觉呈现范式中视觉知觉的神经关联

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349766

Yonghong Huang, K. Hild, M. Pavel, S. Mathan, Deniz Erdoğmuş

{"title":"Neural correlates of visual perception in rapid serial visual presentation paradigms","authors":"Yonghong Huang, K. Hild, M. Pavel, S. Mathan, Deniz Erdoğmuş","doi":"10.1109/MLSP.2012.6349766","DOIUrl":"https://doi.org/10.1109/MLSP.2012.6349766","url":null,"abstract":"Human brain signals associated with visual perceptual processes have been used for image recognition. This paper presents several insights on the neural correlates of human visual perception by analyzing the neural correlates that result when humans view realistic images using a rapid serial visual presentation (RSVP) image display paradigm. We propose an image information extraction model and examine the relationship between the brain evoked response - using event related potential (ERP) characteristics - and the level of difficulty for humans to detect targets as a function of both visual stimulus complexity and task difficulty. We develop a computational model to quantify subject performance and the difficulty of realistic stimuli. Our results show that: (1) more difficult trials produce less prominent ERP patterns, thus reducing the performance of machine-based ERP detection; (2) on average for the same behavioral performance level, a pair of ERP's extracted from two easy trials are more similar than a pair of ERP's from two hard trials; and (3) both stimulus and task difficulty are correlated with neural activity. Our findings indicate that, for dynamic tasks involved in visual information processing, the brain may allocate additional cognitive resources, such as attention, to a given visual stimulus, as the task and/or stimulus difficulty increases.","PeriodicalId":262601,"journal":{"name":"2012 IEEE International Workshop on Machine Learning for Signal Processing","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114690248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 2

Complex-valued analysis and visualization of fMRI data for event-related and block-design paradigms 事件相关和块设计范例的fMRI数据的复杂值分析和可视化

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349790

Pedro A. Rodriguez, T. Adalı, V. Calhoun

引用次数: 0

Voice analysis of patients with neurological disorders using acoustical and nonlinear tools 使用声学和非线性工具分析神经系统疾病患者的声音

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349803

M. E. Dajer, P. Scalassara, J. Marrara, J.C. Pereira

引用次数: 2

Recursive outlier-robust filtering and smoothing for nonlinear systems using the multivariate student-t distribution 基于多元student-t分布的非线性系统递归离群鲁棒滤波与平滑

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349794

R. Piché, S. Särkkä, Jouni Hartikainen

引用次数: 147

Accelerometer based gesture recognition system using distance metric learning for nearest neighbour classification 基于加速度计的手势识别系统，使用距离度量学习进行最近邻分类

2012 IEEE International Workshop on Machine Learning for Signal Processing Pub Date : 2012-11-12 DOI: 10.1109/MLSP.2012.6349717

T. Marasovic, V. Papić

{"title":"Accelerometer based gesture recognition system using distance metric learning for nearest neighbour classification","authors":"T. Marasovic, V. Papić","doi":"10.1109/MLSP.2012.6349717","DOIUrl":"https://doi.org/10.1109/MLSP.2012.6349717","url":null,"abstract":"The need to improve communication between humans and computers has been motivation for defining new communication models, and accordingly, new ways of interacting with machines. In many applications today, user interaction is moving away from traditional keyboards and mouses and is becoming much more physical, pervasive and intuitive. This paper examines hand gestures as an alternative or supplementary input modality for mobile devices. A new gesture recognition system based on the use of acceleration sensor, that is nowadays being featured in a growing number of consumer electronic devices, is presented. Accelerometer sensor readings can be used for detection of hand movements and their classification into previously trained gestures. The proposed system utilizes Mahalanobis distance metric learning to improve the accuracy of nearest neighbour classification. In the approach we adopted, the objective function for metric learning is convex and, therefore, the required optimization can be cast as an instance of semidefinite programming. The experiments, carried out to evaluate system performance, demonstrate its efficacy.","PeriodicalId":262601,"journal":{"name":"2012 IEEE International Workshop on Machine Learning for Signal Processing","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116754416","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 14