Taiyi Zhang

Object tracking using an adaptive Kalman filter combined with mean shift

An object tracking algorithm using an adaptive Kalman filter (KF) combined with mean shift (MS) is proposed. First, the system model of KF is constructed, then the center of the object predicted by KF is used as the initial value of the MS algorithm. The searching result of MS is fed back as the measurement of the adaptive KF, and the estimate parameters of KF are adjusted by the Bhattacharyya coefficient adaptively. The proposed method has the robust ability to track a moving object in consecutive frames under certain real-world complex situations, such as a moving object disappearing partially or totally due to occlusion, fast moving objects, and sudden changes in velocity of a moving object. The experimental results demonstrate that the proposed tracking algorithm is robust and practical.

MODELING OF CHAOTIC SYSTEMS WITH MULTIWAVELET TRANSFORM COMBINED WITH RECURRENT LEAST SQUARES SUPPORT VECTOR MACHINES

A new algorithm for modeling of chaotic systems is presented in this paper. First, more information is acquired utilizing the reconstructed embedding phase space, and the multiwavelets transform provides a sensible decomposition of the data so that the underlying temporal structures of the original time series become more tractable. Second, based on the Recurrent Least Squares Support Vector Machines (RLS-SVM), modeling of the chaotic system is realized. To demonstrate the effectiveness of our algorithm, we use the power spectrum and dynamic invariants involving the Lyapunov exponents and the correlation dimension as criterions, and then apply our method to Chuas circuit time series. The similarity of dynamic invariants between the original and generated time series shows that the proposed method can capture the dynamics of the chaotic time series more effectively.

Predicting nonstationary time series with multi-scale gaussian processes model

The Gaussian processes (GP) model has been successfully applied to the prediction of nonstationary time series. Due to the models covariance function containing an undetermined hyperparameters, to find its maximum likelihood values one usually suffers from either susceptibility to initial conditions or large computational cost. To overcome the pitfalls mentioned above, at the same time to acquire better prediction performance, a novel multi-scale Gaussian processes (MGP) model is proposed in this paper. In the MGP model, the covariance function is constructed by a scaling function with its different dilations and translations, ensuring that the optimal value of the hyperparameter is easy to determine. Although some more time is spent on the calculation of covariance function, MGP takes much less time to determine hyperparameter. Therefore, the total training time of MGP is competitive to GP. Experiments demonstrate the prediction performance of MGP is better than GP. Moreover, the experiments also show that the performance of MGP and support vector machine (SVM) is comparable. They give better performance compared to the radial basis function (RBF) networks.

Modelling of chaotic systems with recurrent least squares support vector machines combined with reconstructed embedding phase space

A new strategy of modelling of chaotic systems is presented. First, more information is acquired utilizing the reconstructed embedding phase space. Then, based on the Recurrent Least Squares Support Vector Machines (RLS-SVM), modelling of the chaotic system is realized. We use the power spectrum and dynamic invariants involving the Lyapunov exponents and the correlation dimension as criterions, and then apply our method to the Chua‘s circuit time series. The simulation of dynamic invariants between the origin and generated time series shows that the proposed method can capture the dynamics of the chaotic time series effectively.

Video object segmentation based on mixtures of probabilistic principal component analysis

A novel video object segmentation algorithm is proposed based on mixtures of probabilistic principal component analysis (MPPCA) in this paper. The number of mixture components of MPPCA is estimated and the expectation maximization (EM) algorithm is initialized through segmentation projection after extracting feature. Then the EM algorithm is applied to estimate the distribution of feature vectors. Finally the segmentation is carried out by clustering each pixel into appropriate component according to maximum likelihood criterion. The proposed algorithm can greatly accelerate the convergence of the EM algorithm since the initial value approximates its real value. As a result, the speed of the video object segmentation is improved. Experimental results have demonstrated that the proposed method can extract moving objects from video sequences successfully. At the same time, the algorithm proposed is more stable.

Support Vector Machine Based Multiresolution Signal Approximation

Multiresolution signal approximation (MSA) provides a simple hierarchical approximation of the signals. And support vector machine (SVM) has been introduced as a novel tool for solving approximation problems. Based on the fact that scale subspaces onto which MSA projects the signals are reproducing kernel Hilbert spaces (RKHS), we integrate the approximation criterion of SVM into MSA and then an SVM based MSA (S-MSA) algorithm is proposed. Experiments exhibit that S-MSA owns better approximation accuracy and smoothness than MSA. Furthermore, quantitative comparison with MSA illustrates the robustness of S-MSA when noises are present

Music style classification with a novel bayesian model

Music style classification by mean of computers is very useful to music indexing, content-based music retrieval and other multimedia applications. This paper presents a new method for music style classification with a novel Bayesian-inference-based decision tree (BDT) model. A database of total 320 music staffs collected from CDs and the Internet is used for the experiment. For classification three features including the number of sharp octave (NSO), the number of simple meters (NSM), and the music playing speed (MPS) are extracted. Following that, acomparative evaluation between BDT and traditional decision tree (DT) model is carried out on the database. The results show that the classification accuracy rate of BDT far superior to existing DT model.