Structured Sparse Coding With the Group Log-regularizer for Key Frame Extraction

Key frame extraction based on sparse coding can reduce the redundancy of continuous frames and concisely express the entire video.However,how to develop a key frame extraction algorithm that can automatically extract a few frames with a low reconstruction error remains a challenge.In this paper,we propose a novel model of structured sparse-codingbased key frame extraction,wherein a nonconvex group log-regularizer is used with strong sparsity and a low reconstruction error.To automatically extract key frames,a decomposition scheme is designed to separate the sparse coefficient matrix by rows.The rows enforced by the nonconvex group log-regularizer become zero or nonzero,leading to the learning of the structured sparse coefficient matrix.To solve the nonconvex problems due to the log-regularizer,the difference of convex algorithm(DCA)is employed to decompose the log-regularizer into the difference of two convex functions related to the l1 norm,which can be directly obtained through the proximal operator.Therefore,an efficient structured sparse coding algorithm with the group log-regularizer for key frame extraction is developed,which can automatically extract a few frames directly from the video to represent the entire video with a low reconstruction error.Experimental results demonstrate that the proposed algorithm can extract more accurate key frames from most Sum Me videos compared to the stateof-the-art methods.Furthermore,the proposed algorithm can obtain a higher compression with a nearly 18% increase compared to sparse modeling representation selection(SMRS)and an 8% increase compared to SC-det on the VSUMM dataset.

Pentamode-Based Coding Metasurface for Underwater Acoustic Stealth

Coding metasurface draws amounts of research interests due to its potential for achieving sophisticated functions in wave manipulation by using simple logical unit cells with out-of-phase responses. In this paper, we present a novel acoustic coding metasurface structure for underwater sound scattering reduction based on pentamode metamaterials. The metasurface is composed of two types of hexagonal pentamode unit cells with phase responses of 0 and π respectively. The units are arranged in random 1-bit coding sequence to achieve low-scattering underwater acoustic stealth effect. Full-wave simulation results are in good accordance with the theoretical expectation. The optimized arrangement resulted in the distribution of scattered underwater acoustic waves and suppression of the far field scattering coefficient over a wide range of incident angles. We show that pentamode-based coding metasurface provides an efficient scheme to achieve underwater acoustic stealth by ultrathin structures.

Protein structural codes and nucleation sites for protein folding

One of the long-standing controversial arguments in protein folding is Levinthal＇s paradox. We have recently proposed a new nucleation hypothesis and shown that the nucleation residues are the most conserved sequences in protein. To avoid the complicated effect of tertiary interactions, we limit our search for structural codes to the nucleation residues. Starting with the hypotheses of secondary structure nucleation and conservation of residues important for folding, we have analysed 762 folds classified as unique by SCOP. Segments of 17 residues around the top 20% conserved amino acids are analysed, resulting in approximately 100 clusters each for the main secondary structure classes of helix, sheet and coil. Helical clusters have the longest correlation range, coils the shortest （four residues）. Strong specific sequence-structure correlation is observed for coil but not for helix and sheet, suggesting a mapping relationship between the sequence and the structure for coil. We propose that the central sequences in these clusters form ＇structural codes＇, a useful basis set for identifying nucleation sites, protein fragments stable in isolation, and secondary structural patterns in proteins （particularly turns and loops）.

New atomic data for Kr XXXV useful in fusion plasma

Energy levels and emission line wavelengths of high-Z materials are useful for impurity diagnostics due to their potential application in the next generation fusion devices.For this purpose,we have calculated the fine structural energies of the 67 levels belonging to the 1s 2,1s2l,1s3l,1s4l,1s5l,and 1s6l configurations of Kr XXXV using GRASP（general purpose relativistic atomic structure package） code.Additionally,we have reported the transition probabilities,oscillator strengths,line strengths,and transition wavelengths for some electric dipole（E1） transitions among these levels.We predict new energy levels and radiative rates,which have not been reported experimentally or theoretically,forming the basis for future experimental work.

Study on the distribution law of random code structure of irregular LDPC codes and its application in eliminating short cycles

The distribution law of the random code structure of randomly constructed irregular low-density parity-check （LDPC） codes is studied, Based on the Progressive Edge-Growth （PEG） algorithm, a new algorithm which can both ellminate short cycles and keep the distribution of the random code structure is presented, The experimentsl results show that the performance of the irregular LDPC codes constructed by the new algorithm is superior to that of the PEG algorithm,

Segment-based coding of color images

Based on the idea of second generation image coding, a novel scheme for coding still images is pre- sented.At first, an image was partitioned with a pulse-coupled neural network; and then an improved chain code and the 2D discrete cosine transform was adopted to encode the shape and the color of its edges respectively.To code its smooth and texture regions, an improved zero-trees strategy based on the 2nd generation wavelet was chosen.After that, the zero-tree chart was selected to rearrange quantified coefficients.And finally some regulations were given according to psychology of various users.Experiments under noiseless channels demonstrate that the proposed method performs better than those of the current one, such as JPEG, CMP, EZW and JPEG2000.

Structural Sparse Representation for Object Detection

Classic sparse representation, as one of prevalent feature learning methods, is successfully applied for different computer vision tasks. However it has some intrinsic defects in object detection. Firstly, how to learn a discriminative dictionary for object detection is a hard problem. Secondly, it is usually very time-consuming to learn dictionary based features in a traditional exhaustive search manner like sliding window. In this paper, we propose a novel feature learning framework for object detection with the structure sparsity constraint and classification error minimization constraint to learn a discriminative dictionary. For improving the efficiency, we just learn sparse representation coefficients from object candidate regions and feed them to a kernelized SVM classifier. Experiments on INRIA Person Dataset and Pascal VOC 2007 challenge dataset clearly demonstrate the effectiveness of the proposed approach compared with two state-of-the-art baselines.

A prototype online database-enabled design framework for wind analysis/design of low-rise buildings

This study presents a development of an advanced cyberbased database-enabled design module for low- rise buildings （DEDM-LR） which provides estimation of the wind-induced responses for main wind force resisting frames by making direct use of pressure time histories measured at a large number of pressure taps over a suite of building models. These responses may be considered in lieu of code-specified load effects in which the overall accuracy may be influenced by the inherent simplifications in codes. In addition, this new automated approach is particularly attractive and advantageous as it allows a web-based online analysis/design via intuitive user-friendly interfaces for both the input and output in terms of familiar web-style forms that are nowadays very common in most of web-based services. Presently, the DEDM-LR hosts an aerodynamic database developed by the Tokyo Polytechnic University （TPU）, Japan for a variety of building configurations like flat, gable, and hip roofs under suburban terrain flow condition with immediate application to other databases. The paper shows the efficacy and validity of the DEDM-LR by walking through its details and examples on selected gable-roofed buildings. The architecture of DEDM-LR platform offers the ability to pool resources by hosting other databases that may become available in the near future.