Multimodality Prediction of Chaotic Time Series with Sparse Hard-Cut EM Learning of the Gaussian Process Mixture Model

The contribution of this work is twofold： （1） a multimodality prediction method of chaotic time series with the Gaussian process mixture （GPM） model is proposed, which employs a divide and conquer strategy. It automatically divides the chaotic time series into multiple modalities with different extrinsic patterns and intrinsic characteristics, and thus can more precisely fit the chaotic time series. （2） An effective sparse hard-cut expec- tation maximization （SHC-EM） learning algorithm for the GPM model is proposed to improve the prediction performance. SHO-EM replaces a large learning sample set with fewer pseudo inputs, accelerating model learning based on these pseudo inputs. Experiments on Lorenz and Chua time series demonstrate that the proposed method yields not only accurate multimodality prediction, but also the prediction confidence interval SHC-EM outperforms the traditional variational 1earning in terms of both prediction accuracy and speed. In addition, SHC-EM is more robust and insusceptible to noise than variational learning.

Two-Dimensional Direction Finding via Sequential Sparse Representations

The problem of two-dimensional direction finding is approached by using a multi-layer Lshaped array. The proposed method is based on two sequential sparse representations,fulfilling respectively the estimation of elevation angles,and azimuth angles. For the estimation of elevation angles,the weighted sub-array smoothing technique for perfect data decorrelation is used to produce a covariance vector suitable for exact sparse representation,related only to the elevation angles. The estimates of elevation angles are then obtained by sparse restoration associated with this elevation angle dependent covariance vector. The estimates of elevation angles are further incorporated with weighted sub-array smoothing to yield a second covariance vector for precise sparse representation related to both elevation angles,and azimuth angles. The estimates of azimuth angles,automatically paired with the estimates of elevation angles,are finally obtained by sparse restoration associated with this latter elevation-azimuth angle related covariance vector. Simulation results are included to illustrate the performance of the proposed method.

Current progress in sparse signal processing applied to radar imaging

Sparse signal processing is a signal processing technique that takes advantage of signal’s sparsity,allowing signal to be recovered with a reduced number of samples.Compressive sensing,a new branch of the sparse signal processing,has become a rapidly growing research field.Sparse microwave imaging introduces the sparse signal processing theory to radar imaging to obtain new theories,new systems and new methodologies of microwave imaging.This paper first summarizes the latest application of sparse microwave imaging,including Synthetic Aperture Radar(SAR),tomographic SAR and inverse SAR.As sparse signal processing keeps evolving,an avalanche of results have been obtained.We also highlight its recent theoretical advances,including structured sparsity,off-grid,Bayesian approaches,and point out new research directions in sparse microwave imaging.

Experimental demonstration of enhanced resolution of a Golay3 sparse-aperture telescope

In this Letter,we report a Golay3 sparse-aperture telescope newly built in the Key Laboratory of Optical Engineering,Chinese Academy of Sciences and present the experimental results of enhanced resolution.The telescope consisting of 3 collector telescopes of 127 mm diameter can achieve a theoretical resolution corresponding to a monolithic aperture of 245 mm diameter.It is shown by the experimental results that the resolution is improved to 3.33μrad with respect to the diffraction limit of 6.07μrad for a single aperture using the Rayleigh criteria at 632 nm.The compact optical configuration and cophasing approach are also described.