We present a novel approach to restoring images blurred by rotational motions, without experiencing geometric coordinate transformations as in traditional restoration. The space-variant blur is decomposed into a serie...
详细信息
We present a novel approach to restoring images blurred by rotational motions, without experiencing geometric coordinate transformations as in traditional restoration. The space-variant blur is decomposed into a series of space-invariant blurs along the blurring paths. By incorporating Bresenham's algorithm into our work, the blurred gray values of the discrete pixels can be fetched along the blurring paths in real time. Thus, the space-variant blur can be quickly removed along the blurring paths. We apply a two-stage process to restore the rectangular blurred image, which results in the proposal of two corresponding restoration algorithms. One removes the blur by deconvolution along the blurring paths, which are completely inside the rectangular image. The other is used in the case when only some of the pixels of some blurring paths are inside the rectangular image, so based on a neighborhood knowledge guide, the information of these pixels is restored with the least cost in terms of the constrained optimization estimation theory. Furthermore, these two restoration algorithms avoid iteration calculations and some time-consuming operations. To determine the blur center and the blur extents from the blurred image in a case of not knowing the rotational motion parameters, we present, based on cross correlation, an effective blur identification method, which becomes an integral part of the proposed approach. The experimental results demonstrate the efficiency of the proposed restoration algorithms and the effectiveness of the blur identification method. (C) 2003 Society of Photo-Optical Instrumentation Engineers.
Transfer learning utilizes data or knowledge in one problem to help solve a related problem. It is particularly useful in electroencephalogram (EEG)-based motor imagery (MI) classification, to handle high intrasubject...
详细信息
Transfer learning utilizes data or knowledge in one problem to help solve a related problem. It is particularly useful in electroencephalogram (EEG)-based motor imagery (MI) classification, to handle high intrasubject and/or cross-subject variations. This article considers offline unsupervised cross-subject MI classification, i.e., we have labeled EEG trials from several source subjects, but only unlabeled EEG trials from the target subject. Existing transfer learning approaches usually make use of the source-domain data directly in target model learning. To protect the privacy of the source subjects, we propose lightweight source-free transfer (LSFT), which first generates source models locally and encapsulates them as model application programming interfaces (APIs), then constructs a virtual intermediate domain to transfer the knowledge in the source domains to the target domain, and finally performs feature adaptation learning. Compared with the existing deep transfer learning approaches, LSFT does not need to transfer from massive source data or models, is computationally efficient, and has a small number of parameters. Experiments on four benchmark MI data sets demonstrated that LSFT outperformed 13 different approaches, including several state-of-the-art transfer learning approaches that make use of the source-domain samples or model parameters directly.
Transfer learning, which utilizes labeled source domains to facilitate the learning in a target model, is effective in alleviating high intra- and inter-subject variations in electroencephalogram (EEG) based brain-com...
详细信息
Transfer learning, which utilizes labeled source domains to facilitate the learning in a target model, is effective in alleviating high intra- and inter-subject variations in electroencephalogram (EEG) based brain-computer interfaces (BCIs). Existing transfer learning approaches usually use the source subjects' EEG data directly, leading to privacy concerns. This paper considers a decentralized privacy-preserving transfer learning scenario: there are multiple source subjects, whose data and computations are kept local, and only the parameters or predictions of their pre-trained models can be accessed for privacy-protection;then, how to perform effective cross-subject transfer for a new subject with unlabeled EEG trials? We propose an offline unsupervised multi-source decentralized transfer (MSDT) approach, which first generates a pre-trained model from each source subject, and then performs decentralized transfer using the source model parameters (in gray-box settings) or predictions (in black-box settings). Experiments on two datasets from two BCI paradigms, motor imagery and affective BCI, demonstrated that MSDT outperformed several existing approaches, which do not consider privacy-protection at all. In other words, MSDT achieved both high privacy-protection and better classification performance.
In this paper,an iterative regularized super resolution (SR) algorithm considering non-Gaussian noise is *** on the assumption of a generalized Gaussian distribution for the contaminating noise,an lp norm is adopted t...
详细信息
In this paper,an iterative regularized super resolution (SR) algorithm considering non-Gaussian noise is *** on the assumption of a generalized Gaussian distribution for the contaminating noise,an lp norm is adopted to measure the data fidelity term in the cost *** the meantime,a regularization functional defined in terms of the desired high resolution (HR) image is employed,which allows for the simultaneous determination of its value and the partly reconstructed image at each iteration *** convergence is thoroughly *** results show the effectiveness of the proposed algorithm as well as its superiority to conventional SR methods.
Acquisition of labeled training samples for affective computing is usually costly and time-consuming, as affects are intrinsically subjective, subtle and uncertain, and hence multiple human assessors are needed to eva...
详细信息
Acquisition of labeled training samples for affective computing is usually costly and time-consuming, as affects are intrinsically subjective, subtle and uncertain, and hence multiple human assessors are needed to evaluate each affective sample. Particularly, for affect estimation in the 3D space of valence, arousal and dominance, each assessor has to perform the evaluations in three dimensions, which makes the labeling problem even more challenging. Many sophisticated machine learning approaches have been proposed to reduce the data labeling requirement in various other domains, but so far few have considered affective computing. This paper proposes two multi-task active learning for regression approaches, which select the most beneficial samples to label, by considering the three affect primitives simultaneously. Experimental results on the VAM corpus demonstrated that our optimal sample selection approaches can result in better estimation performance than random selection and several traditional single-task active learning approaches. Thus, they can help alleviate the data labeling problem in affective computing, i.e., better estimation performance can be obtained from fewer labeling queries.
This paper presented a control design methodology for a proton exchange membrane fuel cell (PEMFC) generation system for residential applications. The dynamic behavior of the generation system is complex in such appli...
详细信息
This paper presented a control design methodology for a proton exchange membrane fuel cell (PEMFC) generation system for residential applications. The dynamic behavior of the generation system is complex in such applications. A comprehensive control design is very important for achieving a steady system operation and efficiency. The control strategy for a 60 kW generation system was proposed and tested based on the system dynamic model. A two-variable single neuron proportional-integral (PI) decoupling controller was developed for anode pressure and humidity by adjusting the hydrogen flow and water injection. A similar controller was developed for cathode pressure and humidity by adjusting the exhaust flow and water injection. The desired oxygen excess ratio was kept by a feedback controller based on the load current. An optimal seeking controller was used to trace the unique optimal power point. Two negative feedback controllers were used to provide AC power and a suitable voltage for residential loads by a power conditioning unit. control simulation tests showed that 60 kW PEMFC generation system responded well for computer-simulated step changes in the load power demand. This control methodology for a 60 kW PEMFC generation system would be a competitive solution for system level designs such as parameter design, performance analysis, and online optimization.
This paper proposed a novel evolutionary template-matching algorithm and studied its convergence problem. The method regarded the image matching as a global optimization problem where the main task is to find the para...
详细信息
This paper proposed a novel evolutionary template-matching algorithm and studied its convergence problem. The method regarded the image matching as a global optimization problem where the main task is to find the parameters of the affine transformation. It can greatly decrease computation amount and quickly detect the affine transformed object from noises-polluted images. Experimental results demonstrate this approach's feasibility and its potential in practical applications.
The 3-D modeling of heads by using optical triangulation techniques is of great interest in the context of virtual reality, telecommunication and computer animation. This paper presents a structured light-based system...
详细信息
The 3-D modeling of heads by using optical triangulation techniques is of great interest in the context of virtual reality, telecommunication and computer animation. This paper presents a structured light-based system mainly for human heads. It is named "3-D Laser Color Scanner" (3DLCS). A 3-D model is obtained with a cylindrical scan. The laser beam is switched on and off using a "light valve" and two successive CCD frames are captured, one with the laser line showing and one without. We can simplify the laser line extracting by subtracting these two images. In this system, two CCD cameras are used to avoid occlusion problems. Color information is read from the CCD when the laser light is absent. Since traditional laser scanner will miss the range data in the low-reflectance areas such as the hair area of human bead, a shape from silhouette algorithm is presented to overcome this problem. Finally, we give some results using our system. The resulting model is suitable for many applications. (C) 2003 Elsevier Ltd. All rights reserved.
Recent methods based on mid-level visual concepts have shown promising capability in human action recognition field. Automatically discovering semantic entities such as parts for an action class remains challenging. I...
详细信息
It is necessary to study the radiation characteristic of metal solid objects for millimeter wave passive guidance. On basis of discussing the grounded theory, the antenna temperature contrast formula of metal solid ob...
详细信息
暂无评论