Past decades, numerous spectral analysis based algorithms have been proposed for dimensionality reduction, which plays an important role in machine learning and artificial intelligence. However, most of these existing...
详细信息
Past decades, numerous spectral analysis based algorithms have been proposed for dimensionality reduction, which plays an important role in machine learning and artificial intelligence. However, most of these existing algorithms are developed intuitively and pragmatically, i.e., on the base of the experience and knowledge of experts for their own purposes. Therefore, it will be more informative to provide some a systematic framework for understanding the common properties and intrinsic differences in the algorithms. In this paper, we propose such a framework, i.e., ldquopatch alignmentrdquo, which consists of two stages: part optimization and whole alignment. With the proposed framework, various algorithms including the conventional linear algorithms and the manifold learning algorithms are reformulated into a unified form, which gives us some new understandings on these algorithms.
The prevalence of Alzheimer's disease (AD) is rising alarmingly as the average age of our population increases. There is no treatment to halt or slow the pathology responsible for AD, however, new drugs are promis...
The prevalence of Alzheimer's disease (AD) is rising alarmingly as the average age of our population increases. There is no treatment to halt or slow the pathology responsible for AD, however, new drugs are promising to reduce the rate of progression. On the other hand, the efficacy of these new medications critically depends on our ability to diagnose AD at the earliest stage. Currently AD is diagnosed through longitudinal clinical evaluations, which are available only at specialized dementia clinics, hence beyond financial and geographic reach of most patients. Automated diagnosis tools that can be made available to community hospitals would therefore be very beneficial. To that end, we have previously shown that the event related potentials obtained from different scalp locations can be effectively used for early diagnosis of AD using an ensemble of classifiers based decision fusion approach. In this study, we expand our data fusion approach to include MRI based measures of regional brain atrophy. Our initial results indicate that ERPs and MRI carry complementary information, and the combination of these heterogeneous data sources using a decision fusion approach can significantly improve diagnostic accuracy.
Short-term forecasting of travel time is essential for the success of intelligent transportation system. In this paper, we review the state-of-art of short-term traffic forecasting models and outline their basic ideas...
详细信息
Short-term forecasting of travel time is essential for the success of intelligent transportation system. In this paper, we review the state-of-art of short-term traffic forecasting models and outline their basic ideas, related works, advantages and disadvantages of each model. An improved adaptive exponential smoothing (IAES) model is also proposed to overcome the drawbacks of the previous adaptive exponential smoothing model. Then, comparing experiments are carried out under normal traffic condition and abnormal traffic condition to evaluate the performance of four main branches of forecasting models on direct travel time data obtained by license plate matching (LPM). The results of experiments show each model seems to have its own strength and weakness. The forecasting performance of IASE is superior to other models in shorter forecasting horizon (one and two step forecasting) and the IASE is capable of dealing with all kind of traffic conditions.
A robust face pose estimation approach is proposed by using face shape statistical model approach and pose parameters are represented by trigonometric functions. The face shape statistical model is firstly built by an...
详细信息
A robust face pose estimation approach is proposed by using face shape statistical model approach and pose parameters are represented by trigonometric functions. The face shape statistical model is firstly built by analyzing the face shapes from different people under varying poses. The shape alignment is vital in the process of building the statistical model. Then, six trigonometric functions are employed to represent the face pose parameters. Lastly, the mapping function is constructed between face image and face pose by linearly relating different parameters. The proposed approach is able to estimate different face poses using a few face training samples. Experimental results are provided to demonstrate its efficiency and accuracy.
A new method to reconstruct 3D scene points from nonparallel stereo is proposed. From a pair of conjugate images in an arbitrarily configured stereo system that has been calibrated, coordinates of 3D scene points can ...
详细信息
A new method to reconstruct 3D scene points from nonparallel stereo is proposed. From a pair of conjugate images in an arbitrarily configured stereo system that has been calibrated, coordinates of 3D scene points can be computed directly using the method, bypassing the process of rectifying images or iterative solution involved in existing methods. Experiment results from both simulated data and real images validate the method. Practical application to surgical navigator shows that the method has advantages to improve efficiency and accuracy of 3D reconstruction from nonparallel stereo system in comparison with the conventional method that employs algorithm for standard parallel axes stereo geometry.
A new model-based speech enhancement algorithm by variational Bayesian learning was proposed in this paper. Autoregressive process was used to model speech signal and its order was determined automatically. Clean spee...
详细信息
A new model-based speech enhancement algorithm by variational Bayesian learning was proposed in this paper. Autoregressive process was used to model speech signal and its order was determined automatically. Clean speech signal could be estimated using a variational Kalman smoother. Moreover, overfitting was avoided in the learning of model parameter and model structure. Experimental results compared with Kalman filter-based enhancement and spectral subtraction methods demonstrate the performance of our algorithm.
In order to make full use of heterogeneous multi-sensor data to serve urban intelligent transportation systems, a real-time urban traffic state fusion model was proposed, named federated evidence fusion model. The mod...
详细信息
In order to make full use of heterogeneous multi-sensor data to serve urban intelligent transportation systems, a real-time urban traffic state fusion model was proposed, named federated evidence fusion model. The model improves conventional D-S evidence theory in temporal domain, such that it can satisfy the requirement of real-time processing and utilize traffic detection information more efficaciously. The model frame and computation al procedures are given. In addition, a generalized reliability weight matrix of evidence is also presented to increase the accuracy of estimation. After that, a simulation test is presented to explain the advantage of the proposed method in comparison with conventional D-S evidence theory. Besides, the validity of the model is proven by the use of the data of loop detectors and GPS probe vehicles collected from an urban link in Shanghai. Results of the experiment show that the proposed approach can well embody and track traffic state at character level in real-time conditions.
In order to perform a high-quality interactive rendering of large medical data sets on a single off-the-shelf PC, a LOD selection algorithm for multi-resolution volume rendering using 3D texture mapping is presented, ...
详细信息
In order to perform a high-quality interactive rendering of large medical data sets on a single off-the-shelf PC, a LOD selection algorithm for multi-resolution volume rendering using 3D texture mapping is presented, which uses an adaptive scheme that renders the volume in a region-of-interest at a high resolution and the volume away from this region at lower resolutions. The algorithm is based on several important criteria, and rendering is done adaptively by selecting high-resolution cells close to a center of attention and low-resolution cells away from this area. In addition, our hierarchical level-of-detail representation guarantees consistent interpolation between different resolution levels. Experiments have been applied to a number of large medical data and have produced high quality images at interactive frame rates using standard PC hardware.
In this paper,three dimensions kinematics and kinetics simulation are discussed for hardware realization of a physical biped walking-chair *** direct and inverse close-form kinematics solution of the biped walking-cha...
详细信息
In this paper,three dimensions kinematics and kinetics simulation are discussed for hardware realization of a physical biped walking-chair *** direct and inverse close-form kinematics solution of the biped walking-chair robot is *** gaits are realized with the kinematics solution,including walking straight on level floor,going up stair,squatting down and standing *** Moment Point(ZMP)equation is analyzed considering the movement of the *** simulated biped walking-chair robot is used for mechanical design,gaits development and validation before they are tested on real robot.
Since in most blind source separation(BSS)algorithms the estimations of probability density function(pdf)of sources are fixed or can only switch between one sup-Gaussian and other sub-Gaussian model,they may not be ef...
详细信息
Since in most blind source separation(BSS)algorithms the estimations of probability density function(pdf)of sources are fixed or can only switch between one sup-Gaussian and other sub-Gaussian model,they may not be efficient to separate sources with different *** to solve the problem of pdf mismatch and the separation of hybrid mixture in BSS,the generalized Gaussian model(GGM)is introduced to model the pdf of the sources since it can provide a general structure of univariate *** great advantage is that only one parameter needs to be determined in modeling the pdf of different sources,so it is less complex than Gaussian mixture *** using maximum likelihood(ML)approach,the convergence of the proposed algorithm is *** computer simulations show that it is more efficient and valid than conventional methods with fixed pdf estimation.
暂无评论