Search

In the deep learning-based video action recognitio, the function of the neural network is to acquire spatial information, motion information, and the associated information of the above two kinds of information over an uneven time span. This paper puts forward a network extracting video sequence semantic information based on deep integration of local Spatial-Temporal information. The network uses 2D Convolutional Neural Network (2DCNN) and Multi Spatial-Temporal scale 3D Convolutional Neural Network (MST_3DCNN) respectively to extract spatial information and motion information. Spatial information and motion information of the same time quantum receive 3D convolutional integration to generate the temporary Spatial-Temporal information of a certain moment. Then, the Spatial-Temporal information of multiple single moments enters Temporal Pyramid Net (TPN) to generate the local Spatial-Temporal information of multiple time scales. Finally, bidirectional recurrent neutral network is used to act on the Spatial-Temporal information of all parts so as to acquire the context information spanning the length of the entire video, which endows the network with video context information extraction capability. Through the experiments on the three video action recognitio common experimental data sets UCF101, UCF11, UCFSports, the Spatial-Temporal information deep fusion network proposed in this paper has a high correct recognition rate in the task of video action recognitio.

Self-supervised learning can be adopted to mine deep semantic information of visual data without a large number of human-annotated supervision by using a pretext task to pretrain a model. In this study, we proposed a novel self-supervised learning paradigm, namely multi-task self-supervised (MTSS) representation learning. Unlike existing self-supervised learning methods, which pretrain neural networks on the pretext task and then fine-tune the parameters of neural networks on the downstream task, in our scheme, downstream and pretext tasks are considered primary and auxiliary tasks, respectively, and are trained simultaneously. Our method involves maximizing the similarity of two augmented views of an image as an auxiliary task and using a multi-task network to train the primary task alongside the auxiliary task. We evaluated the proposed method on standard datasets and backbones through a rigorous experimental procedure. Experimental results revealed that proposed MTSS can achieve better performance and robustness than other self-supervised learning methods on multiple image classification data sets without using negative sample pairs and large batches. This simple yet effective method can inspire people to rethink self-supervised learning.

In video action recognition based on deep learning, the design of the neural network is focused on how to acquire effective spatial information and motion information quickly. This paper proposes a kind of deep network that can obtain both spatial information and motion information in video classification. It is called MDFs (the multidimensional motion features of deep feature map net). This method can be used to obtain spatial information and motion information in videos only by importing image frame data into a neural network. MDFs originate from the definition of 3D convolution. Multiple 3D convolution kernels with different information focuses are used to act on depth feature maps so as to obtain effective motion information at both spatial and temporal. On the other hand, we split the 3D convolution at space dimension and time dimension, and the spatial network feature map has reduced the dimensions of the original frame image data, which realizes the mitigation of computing resources of the multichannel grouped 3D convolutional network. In order to realize the region weight differentiation of spatial features, a spatial feature weighted pooling layer based on the spatial-temporal motion information guide is introduced to realize the attention to high recognition information. By means of multilevel LSTM, we realize the fusion between global semantic information acquisition and depth features at different levels so that the fully connected layers with rich classification information can provide frame attention mechanism for the spatial information layer. MDFs need only to act on RGB images. Through experiments on three universal experimental datasets of action recognition, UCF10, UCF11, and HMDB51, it is concluded that the MDF network can achieve an accuracy comparable to two streams (RGB and optical flow) that requires the import of both frame data and optical flow data in video classification tasks.

The aim of the present study was to explore the safety of apatinib plus S-1 in treating advanced solid tumors after failure of two or more lines of chemotherapy. A total of 33 patients with advanced cancer treated between April 2016 to March 2019 were retrospectively analyzed. Of these, 13 patients had non-small cell lung cancer (NSCLC), 13 patients had SCLC, 4 patients had esophageal cancer and 3 had cervical cancer. All patients were treated with apatinib 250 mg once daily combined with S-1 60 mg/m2 twice daily for 14 days, repeated every 3 weeks. Adverse reactions were observed until aggravation of adverse reactions beyond the tolerable range or disease progression, and the survival rate and clinical benefits were calculated. The results suggested that the incidence rate of adverse effects (grade 3-4) was 45.5% (15/33). The top three severe adverse effects were hypertension (15.2%), thrombocytopenia (12.1%) and proteinuria (9.1%). A total of 2 patients with lung squamous-cell carcinomas died of severe pulmonary hemorrhage. Other adverse reactions were tolerated in the cohort. A total of 10 patients achieved partial response and the objective response rate was 30.3%. Furthermore, 13 patients achieved stable disease and 10 patients had progressive disease, and accordingly, the disease control rate was 72.7%. In conclusion, apatinib plus S-1 for advanced solid tumor patients as palliative treatment have a certain efficacy and was relatively safe but should be used with caution in patients with squamous-cell lung carcinoma and the efficacy and safety requires further assessment.

Stock return prediction can help investors make better investment decisions and trends of country's economics. However, most of methods for stock return prediction are based on time-series models, treating the stocks as independent from each other. Inter-relations among stocks' time series are out of consideration. In this work, a Multi-Channel Temporal Graph Convolutional Neural Network (MCT-GCN) is proposed to optimize stock movement prediction. Experiments show that its performance is greater than benchmark algorithms, LSTM in the S&P 500.

Purpose To evaluate the safety and efficacy of Cyberknife® (CK) for the treatment of primary or recurring thymic tumours. Materials and methods We retrospectively reviewed 12 patients (16 tumour lesions) with primary or recurring thymic tumours who were treated with CK between March 2008 and October 2017. Their data was stored in prospectively collected database. Kaplan–Meier method was used to calculate survival curves. Results Five patients (41.7%), who had inoperable disease or refused surgery, were treated with CK initially, and 7 patients (58.3%) were treated with CK when they had recurrence diseases. The disease sites treated with CK were primary tumour site (5), regional lymph nodes (4), tumour bed (3), chest wall (2), pleura (1), and bone (1). The median target volume was 43.8 cm3 (range, 13.1–302.5 cm3) for the 16 tumour lesions. The median follow-up time was 69.3 months (range, 9.7–124.8 months). The median survival time was 48.2 months, and the 5-year and 10-year OS rates were 68.2% and 45.5%, respectively. A high response rate for the tumour lesions irradiated with CK was obtained. Only one patient (8%) experienced in-field recurrence, and the 5-year local recurrence free survival was 90.9%. A case indicated that CK may induce the abscopal effect, which provides the potential to combine CK and immunotherapy. No severe radiation related toxicities were observed, and no treatment related death occurred. Conclusion CK treatment resulted in good outcomes, particularly local control, with minimal side effects, in highly selected patients with primary and recurring thymic tumours. More studies with larger sample are needed.

Jing Luo,1–3 Sainan Hu,4 Tingting Wei,1–3 Jifeng Sun,1–3 Ningbo Liu,1–3 Jun Wang1–3 1Department of Radiation Oncology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, 2Key Laboratory of Cancer Prevention and Therapy, 3Tianjin’s Clinical Research Center for Cancer, Tianjin 300060, China; 4Jiangsu Cancer Hospital, Jiangsu Institute of Cancer Research, The Affiliated Cancer Hospital of Nanjing Medical University, Nanjing 210000, China Purpose: Plasma TGF-β1 protein levels reportedly may predict the treatment outcomes of lung cancer. We hypothesized that in patients with lung cancer treated with radiation therapy (RT), TGF-β1 levels may correlate with the percentages of CD4+ T cells, CD8+ T cells, and the CD4+/CD8+ T cell ratio in peripheral blood. Patients and methods: Eighty-two lung cancer patients satisfied the inclusion criteria. Platelet-poor plasma was obtained before RT, at the second and fourth weeks during RT, and at the end of RT (pre-, during-, and post-RT, respectively). TGF-β1 was measured via ELISA, while recording the percentages of lymphocyte subsets in peripheral blood. Short-term efficacy was categorized as complete response, partial response, stable disease, or progressive disease. Results: Patients who had significantly lower TGF-β1 protein levels after RT than pre-RT seemed to have a better short-term effect (P

Multiactivity 3-D pose tracking is an extremely difficult computer vision task due to various activity types and high-dimensional state space. To solve these challenges, a novel generative framework which incorporates separate and unified motion models into one system is proposed. On one hand, pose tracking benefits from the representations of activities in low-dimensional space, but the complexity of the representations increases with the number of activities. Obviously, the separate modeling of different activities can improve the scalability of models. On the other hand, there often exist similar activities in a long motion sequence, they should be grouped and modeled in a unified model, which can enhance human pose tracking stability and reduce the optimization process over motion model for a large number of actions. Our incorporated model not only combines their advantages but also overcomes their disadvantages. Finally, a particle-based joint annealing optimization is utilized for tracking. During pose tracking, the distribution of particles depends on the outputs of action recognition system. Thus, the particles are transferred amongst different models and most of them are distributed in the model that corresponds to the currently observed action. To propagate the particles among motion models and quickly respond to the transition between different activities, a group of short and smooth transition bridges are constructed so that the interactivity motions seem to be more realistic and natural. Extensive results, via qualitative and quantitative analyses, verify the effectiveness of our model in multiactivity pose tracking.

This paper introduces a new binocular stereo deep learning network based on point cloud, which can realize higher precision point cloud reconstruction through continuous iteration of the network. Our method directly carries out point cloud processing on the target, calculates the difference between the current depth map and the real depth, estimates the loss according to the predicted point cloud and the information of the dual view input image, and then uses the appropriate loss function to iteratively process the point cloud. In addition, we can customize the number of iterations to achieve higher precision point cloud effect. The proposed network basically achieves good results on KITTI data set.

We develop a model based on first principles calculations and experimental data for the thermoelectric properties of p -type CaMg 2 Bi 2 . The thermoelectric performance was assessed and optimized based on the model. The model predicts bipolar reduction of ZT at high temperatures (e.g. > 800 K). This indicates the thermoelectric performance of the material can be further optimized by tuning the carrier concentration beyond the bipolar regime.

Recognizing and tracking multiple activities are all extremely challenging machine vision tasks due to diverse motion types included and high-dimensional (HD) state space. To overcome these difficulties, a novel generative model called composite motion model (CMM) is proposed. This model contains a set of independent, low-dimensional (LD), and activity-specific manifold models that effectively constrain the state search space for 3D human motion recognition and tracking. This separate modeling of activity-specific movements can not only allow each manifold model to be optimized in accordance with only its respective movement, but also improve the scalability of the models. For accurate tracking with our CMM, a particle filter (PF) method is thus employed and then the particles can be distributed in all manifold models at each time step. In addition, an efficient activity switching strategy is proposed to dominate the particle distribution on all LD manifolds. To diffuse the particles amongst manifold models and respond quickly to the sudden changes in the activity, a set of visually-reasonable and kinematically-realistic transition bridges are synthesized by using the good properties of LD latent space and HD observation space, which enables the inter-activity motions seem more natural and realistic. Finally, a pose hypothesis that can best interpret the visual observation is selected and then used to recognize the activity that is currently observed. Extensive experiments, via qualitative and quantitative analyses, verify the effectiveness and robustness of our proposed CMM in the tasks of multi-activity 3D human motion recognition and tracking.

In this study, we use random forest to predict several cryptocurrencies’ prices by using part of factors in Alpha101 [1] to represent features from the history of cryptocurrencies’ market data on Binance and Bitfinex. The result shows our strategy with some factors from Alpha101 is effective in cryptocurrency trading.

A high thermoelectric power factor not only enables a potentially high figure of merit ZT but also leads to a large output power density, and hence it is pivotal to find an effective route to improve the power factor. Previous reports on the manipulation of carrier scattering mechanisms (e.g. ionization scattering) were mainly focused on enhancing the Seebeck coefficient. In contrast, here we demonstrate that by tuning the carrier scattering mechanism in n-type Mg3Sb2-based materials, it is possible to noticeably improve the Hall mobility, from ∼19 to ∼77 cm2 V−1 s−1, and hence substantially increase the power factor by a factor of 3, from ∼5 to ∼15 μW cm−1 K−2. The enhancement in mobility is mainly due to the reason that ionization scattering has been converted into mixed scattering between ionization and acoustic phonon scattering, which less effectively scatters the carriers. The strategy of tuning the carrier scattering mechanism to improve the mobility should be widely applicable to various material systems for achieving better thermoelectric performance.

The structural, electronic and optical properties of AuBX2 (X = S, Se, Te) and AuMTe2 (M = Al, Ga, In) crystals with a chalcopyrite structure were investigated using first principles projector augmented wave (PAW) method with the PBE-GGA functional and the modified Becke-Johnson (mBJ) potential including spin-orbit coupling (SOC) effects. The results show finite band gaps in all the compounds with mBJ and SOC. AuBS2 and AuBSe2 are found to have indirect band gaps, and the tellurides are effective direct band-gap semiconductors. The calculated optical quantites using mBJ and SOC show low energy loss and reflectivity, and high absorption capability in the infrared and visible regions for AuBX2 (X = S, Se, Te) and AuAlTe2, which suggest potentials of using these compounds for solar cell applications.

BACKGROUND: This study aimed to determine the extent to which the survival outcomes of patients with N2-III non-small cell lung cancer (NSCLC) following surgery differ by histological subtype. METHODS: Patients with N2-III NSCLC receiving surgery between 2010 to 2016 were included using the Surveillance, Epidemiology and End Results (SEER) database. Cox proportional hazards models were used to identify risk factors associated with overall survival (OS) and non-cancer mortality. The Kaplan-Meier method with log-rank tests was used to estimate survival. Propensity score matching (PSM) was used. Statistical significance was defined as P

Visual object tracking is a challenging task in computer vision. Recently, due to high efficiency and performance, correlation filter-based trackers have attracted much attention. In this paper, we propose a collaborative correlation filter tracking framework. First, we learn multiple kernelized correlation filters for different features independently, and fuse their response maps to obtain a more reliable response map for translation estimation. Then, we learn a scale correlation filter to handle the scale variation. Moreover, in order to further improve the tracking accuracy, we build an online detector to re-detect objects in local neighbor region, when the tracking result is unreliable. Extensive evaluations on the recent benchmark datasets show that the proposed algorithm performs favorably against several state-of-the-art methods.

In historical documents, Chinese characters can be categorized into more than 8000 categories and are hard to recognize by directly applying classic methods. Furthermore, the lack of well-labelled data makes it difficult to recognize them using deep learning based methods. In this paper, we introduce a historical Chinese text recognizer that is trained by data labelled in page-level without the alignment of each text line. We propose Adaptive Gradient Gate AGG) to reduce the influence of misalignments between text line images and labels. With the help of the AGG, the error rate of the proposed text recognizer can be reduced by over 35%. Furthermore, we find that the implicit language model induced by a text recognition network with Convolutional Neural Networks(CNNs) and Connectionist Temporal Classification (CTC) as one important factor to achieve high recognition performance. Lastly, we compare the proposed recognizer with the three leading optical character recognition systems on the market. Our comparison reveals that the proposed recognizer outperforms other character recognition systems significantly.

Appearance models play an important role in visual tracking. Effective modeling of the appearance of tracked objects is still a challenging problem because of object appearance changes caused by factors, such as partial occlusion, illumination variation and deformation, etc. In this paper, we propose a tracking method based on the patch descriptor and the structural local sparse representation. In our method, the object is firstly divided into multiple non-overlapped patches, and the patch sparse coefficients are obtained by structural local sparse representation. Secondly, each patch is further decomposed into several sub-patches. The patch descriptors are defined as the proportion of sub-patches, of which the reconstruction error is less than the given threshold. Finally, the appearance of an object is modeled by the patch descriptors and the patch sparse coefficients. Furthermore, in order to adapt to appearance changes of an object and alleviate the model drift, an outlier-aware template update scheme is introduced. Experimental results on a large benchmark dataset demonstrate the effectiveness of the proposed method.

In this paper, we present a novel algorithm called robust sparse tensor subspace learning (RSTSL) for 3D human pose regression, and it further extends the latest presented tensor learning to a sparse case. A set of interrelated sparse discriminant projection matrices for feature extraction can be obtained by introducing k-mode optimization and elastic-net algorithm into the objective function of RSTSL and each non-zero element in each discriminant projection matrix is selected from the most typical variables. Thus, the most important low-dimensional tensor feature (LDTF) that corresponds to a test image (i.e., high-order tensor) is extracted through sparse projection transformation. Moreover, we present a novel regression model called optimal-order support tensor regression (OOSTR) to build a finest mapping function between LDTF and 3D human pose configuration. Extensive simulations are conducted on two human motion databases, HumanEva and Brown databases, experimental results show that our proposed RSTSL can not only weaken the sensitivity to incoherent human motions caused by transient occlusion of cameras, sudden change in human velocity and low-frame rate but also strengthen the robustness to silhouette ambiguity, obstacle occlusion and random noise. All the results have confirmed that our tracking system achieves the most significant performance against the compared the state-of-the-art approaches, especially in the complicated human motion databases with different clustered backgrounds, human movements, clothing-style, illumination and subjects like HumanEva database.

Due to the recent advancements in high- throughput sequencing technologies, biomedical research is faced with ever increasing quantities of data, and the storage or transmission of the huge amount of data is one of the concerns. So, we presented a novel hybrid particle swarm optimization based memetic algorithm (HPMA) for DNA sequence compression. In HPMA, within the framework of the memetic algorithm, dynamic comprehensive learning particle swarm optimization method is used for global search, and two adaptive local search operators including center symmetry mutation differential evolution operator and adaptive chaotic search operator work in a cooperative way. HPMA looks for the global optimal code book based on extended approximate repeat vector, by which the DNA sequence will be compressed. Experiments were conducted on 19 high-dimensional functions and 11 real DNA sequences. The results show that HPMA is more competitive in both the performance and scalability, and also attains better compression ability than other representative DNA-specific algorithms on DNA sequence data.

NaSbS2 is a semiconductor that was recently shown to have remarkable efficacy as a solar absorber indicating efficient charge collection even in defected material. We report first principles calculations of properties that show (1) an indirect gap only slightly smaller than the direct gap, which may impede recombination of photoexcited carriers, (2) highly anisotropic electronic and optical properties reflecting a layered crystal structure, (3) a pushed up valence band maximum due to repulsion from the Sb 5s states and (4) cross-gap hybridization between the S p derived valence bands and the Sb 5p states. This latter feature leads to enhanced Born effective charges that can provide local screening and therefore defect tolerance. These features are discussed in relation to the performance of the compound as a semiconductor with efficient charge collection.

This work presents a novel method for human action recognition based on feature level fusion and random projection. The proposed method exploits both spatial-temporal gradient features and Gabor features of the action in video, which helps representing the action more accurately after feature level fusion. Meanwhile, the random projection is employed to reduce the dimensionality of features effectively. In addition, the Bayesian parameter estimation is applied to the Latent Dirichlet Allocation (LDA) topic model. It reflects the action distribution of different videos as well as reduces the complexity of parameter estimation. Experimental results on publically available datasets KTH dataset indicate that the proposed method not only outperforms the single local descriptor approach but also improves the recognition performance compared with the baseline classifier in the same experimental settings.

In this paper, a novel framework based on action temporal-spatial semantic guide for tracking 3D human poses from multi-view images is proposed. The inherent ambiguities caused by the lack of depth information or self-occlusion make it difficult to correctly track 3D human poses from 2D observations. To address this problem, we attempt to capture the important action temporal-spatial semantic knowledge to guide the update of rough 3D human pose hypotheses. The action semantic modeling is implemented by using both temporal and spatial restrictions. The action temporal restriction maintains time-consistency among sequential 3D human movements through a temporal manifold movement template while the action spatial restriction excavates movement correlation coefficient between human body limbs through linear partial least square regression algorithm. Extensive experiments on datasets of various human activities demonstrate that our proposed framework achieves the most significant pose tracking performance compare to the state-of-the-art methods without incorporating action temporal-spatial semantic guide.

In the deep learning-based video action recognition, the function of the neural network is to acquire spatial information, motion information, and the associated information of the above two kinds of information over an uneven time span. We propose a network for extracting semantic information of video sequences based on the deep fusion feature of local spatial–temporal information. Convolutional neural networks (CNNs) are used in the network to extract local spatial information and local motion information, respectively. The spatial information is in three-dimensional convolution with the motion information of the corresponding time to obtain local spatial–temporal information at a certain moment. The local spatial–temporal information is then input into the long- and short-time memory (LSTM) to obtain the context relationship of the local spatial–temporal information in the long-time dimension. We add the ability of the regional attention mechanism of video frames in the neural network mechanism for obtaining context. That is, the last layer of convolutional layer spatial information and the first layer of the fully connected layer are, respectively, input into different LSTM networks, and the outputs of the two LSTMs at each time are merged again. This enables a fully connected layer that is rich in categorical information to provide a frame attention mechanism for the spatial information layer. Through the experiments on the three action recognition common experimental datasets UCF101, UCF11, and UCFSports, the spatial–temporal information deep fusion network proposed has a high correct recognition rate in the task of action recognition.

In this paper, a new algorithm is proposed for the object tracking in the complex background, in which discrete cosine transform (DCT) features of image are used for expressing candidate objects and the templates, casting tracking as a sparse approximation problem. Firstly, the location of object is labeled and templates of DCT features are constructed in the first frame. Secondly, the candidate objects are searched by random particles in the next frame and the sparsity is achieved by solving an h-regularized least-squares problem. The candidate with the smallest object templates projection error is chosen as the tracking object. Finally, the DCT feature templates are updated using the most recent tracking results to capture changes of the object appearance. Experimental results show the effectiveness of the proposed method comparing with other tracking algorithms.

Monocular image-based three-dimensional (3-D) human pose recovery aims to retrieve 3-D poses using the corresponding two-dimensional image features. Therefore, the pose recovery performance highly depends on the image representations. We propose a multispectral embedding-based deep neural network (MSEDNN) to automatically obtain the most discriminative features from multiple deep convolutional neural networks and then embed their penultimate fully connected layers into a low-dimensional manifold. This compact manifold can explore not only the optimum output from multiple deep networks but also the complementary properties of them. Furthermore, the distribution of each hierarchy discriminative manifold is sufficiently smooth so that the training process of our MSEDNN can be effectively implemented only using few labeled data. Our proposed network contains a body joint detector and a human pose regressor that are jointly trained. Extensive experiments conducted on four databases show that our proposed MSEDNN can achieve the best recovery performance compared with the state-of-the-art methods.

Estimating three-dimensional (3D) human poses from a single camera is usually implemented by searching pose candidates with image descriptors. Existing methods usually suppose that the mapping from feature space to pose space is linear, but in fact, their mapping relationship is highly nonlinear, which heavily degrades the performance of 3D pose estimation. We propose a method to recover 3D pose from a silhouette image. It is based on the multiview feature embedding (MFE) and the locality-sensitive autoencoders (LSAEs). On the one hand, we first depict the manifold regularized sparse low-rank approximation for MFE and then the input image is characterized by a fused feature descriptor. On the other hand, both the fused feature and its corresponding 3D pose are separately encoded by LSAEs. A two-layer back-propagation neural network is trained by parameter fine-tuning and then used to map the encoded 2D features to encoded 3D poses. Our LSAE ensures a good preservation of the local topology of data points. Experimental results demonstrate the effectiveness of our proposed method.

Estimating three-dimensional (3-D) pose from a single image is usually performed by retrieving pose candidates with two-dimensional (2-D) features. However, pose retrieval usually relies on the acquisition of sufficient labeled data and suffers from low retrieving accuracy. Acquiring a large amount of unconstrained 2-D images annotated with 3-D poses is difficult. To solve these issues, we propose a coupled-source framework that integrates two independent training sources. The first source contains only 3-D poses, and the second source contains images annotated with 2-D poses. For accurate retrieval, we present a local-topology preserved sparse coding (LTPSC) to generate pose candidates, where the estimated 2-D pose of a test image is regarded as features for pose retrieval and represented as a sparse combination of features in the exemplar database. Our LTPSC can ensure that the semantically similar poses are retrieved with larger probabilities. Extensive experiments validate the effectiveness of our method.

Object tracking is a challenging task in many computer vision applications due to occlusion, scale variation and background clutter, etc. In this paper, we propose a tracking algorithm by combining discriminative global and generative multi-scale local models. In the global model, we teach a classifier with sparse discriminative features to separate the target object from the background based on holistic templates. In the multi-scale local model, the object is represented by multi-scale local sparse representation histograms, which exploit the complementary partial and spatial information of an object across different scales. Finally, a collaborative similarity score of one candidate target is input into a Bayesian inference framework to estimate the target state sequentially during tracking. Experimental results on the various challenging video sequences show that the proposed method performs favorably compared to several state-of-the-art trackers.

This paper proposes a compression algorithm based on K-means clustering for high-through DNA sequence (DNAC-K). In DNAC-K, we create cluster of sequences based on K-means clustering method at first, then iterate clusters according to the edit distances of subsequences, and finally, adopt Huffman coding to encode the result of clustering result. Experimental results on several sequencing data sets demonstrate better performance of DNAC-K than many of the current high-throughput DNA sequence compression algorithms.

A new method of spatial constraints-based maximum likelihood estimation (SC-based MLE) is proposed to process latent variables data of incomplete human motions cycle, and improve the GPDM learning and estimation in this paper, which can make the GPDM learn the samples of incomplete human motions cycle to estimate the new human motions. The proposed method has the GPDM learning less depend on training samples of the complete human motions cycle, and save the training samples. We verify the validity and efficiency of the proposed method, through the experiments of human motions estimation using the samples of incomplete and complete human motions cycle for training respectively.

The recent experimental work shows that perovskite BaSnO3 can be heavily doped by K to become a stable p-type semiconductor. Here, we find that p-type perovskite BaSnO3 retains transparency for visible light while absorbing strongly in the infrared below 1.5 eV. The origin of the remarkable optical transparency even with heavy doping is that the interband transitions that are enabled by empty states at the top of the valence band are concentrated mainly in the energy range from 0.5 to 1.5 eV, i.e., not extending past the near IR. In contrast to n-type, the Burstein-Moss shift is slightly negative, but very small reflecting the heavier valence bands relative to the conduction bands.

Recently, distributed video coding (DVC) has attracted much more attention with advances in low-complexity encoder and flexible architecture, which could be effectively applied in mobile video communication and video surveillance network. But the gap with the conventional video codecs is still large. One of the key factors is the quality of side information as it directly affects the system's rate distortion performance. In this paper, existing algorithms of side information generation in multi-view DVC system are compared, and a new scheme based on feedback channel and fusion will be proposed. Experimental results represent that the proposed scheme can provide better rate distortion performance with the existing ones.

This work presents an algorithm for detecting planar regions in uncalibrated stereo images. The algorithm can also apply to situation that feature point matching between images is unreliable. First, a set of feature points are detected and matched between image pairs. Then an initial homography is computed from match points. Last, Planar region augmentation is performed to enlarge the initial planar region.

In this paper, a new multi-focus image fusion algorithm based on the Nonsubsampled Contourlet Transform (NSCT) is proposed. As a more effective multi-resolution analysis method, NSCT can give a better performance in image processing than the existing methods. Because Human Visual system(HVS) can only be sensitivity to regional characteristic rather than single pixel characteristic, region variance - maximum-based rule and region energy- maximum-based rule are employed to obtain high-frequency and low-frequency coefficients. In the experiments, we compared this strategy with wavelet-based algorithm and contourlet transform -based algorithm. Experimental results show that the fused result based on NSCT can achieve a better performance than the other algorithm.

In this paper, a novel coding scheme of multi-view image is proposed. It is based on fusion rather than redundancy removal since the former is more similar to visual characteristic of people. A multi-view image system using this scheme is designed. BEMD is applied during the fusion and compression process because of its excellent performance on image processing. Experimental results show, without reducing the image quality, the proposed scheme can provide better compression ratio than the previous ones, which open the doors for future applications.

NSCT (Non-Subsampled Contourlet Transform) is widely used in image fusion according to its multiscale, multiorientation and translation invariance. In this paper, a new intermediate view synthesis approach for free viewpoint video based on NSCT and image fusion is proposed. SIFT descriptor is used to detect the distinctive features as matching points of the given images. Then homography matrix is calculated by the matching points, so that the given images can be wrapped to the middle position, and transformed into NSCT domain to fuse, then we fill the holes by blending neighboring colors, the image generated is the intermediate view. The experimental results show that the viability of the proposed method is good.

In this paper, a new image compression scheme is proposed. The self similarity of a static image is analyzed in order to implement image compression on basis of similarity and fusion rather than redundancy removal since the former is more similar to visual characteristic of people. Experimental result shows the scheme presented in the paper can has a potential performance on image and video coding.

As we know, the 802.11a/b/g wireless technology standards can provide a number of non-overlapping channels, and wireless mesh networks can take full advantage of available channels to improve network throughput. In this paper, according to the traffic characteristics of the wireless mesh networks, we use group-based management of multi-channel allocation method, and do some simulation experiments on opnet modeler14.0. The final result shows that it can get better network throughput, compared with using single channel only.

This paper proposed an enhanced algorithm of palmprint recognition. The 2D Gabor was done firstly to filter in the main direction and strengthen the primary line’s information. Then we adopted wavelet transform to decompose the palmprint image, and extract the low frequency component. Two-Dimensional Principal Component Analysis(2DPCA) can avoid transforming from image matrix to 1D vector so as to reduce the computational complexity and gain the eigenvalue of image. However, some noises will affect the algorithm due to the tiny rotation and squeezing in the samples collection. In order to improve the traditional 2DPCA, and increase the recognition rate of palmprints, the paper applied the Moment Invariance. It is not sensitive to the noise mentioned above, and can prevent from being influenced by them. This paper combined the two methods, and calculated the eigenvalue again and again, then matched each other by nearest distance rule. The experiment shows that 2DPCA combining with moment invariances can improve recognition rate compare to 2DPCA.

A low complexity spatial-domain error concealment method is proposed. The edges in lost MB's are detected and recoveried by using neighborhood MB's' gradient data. The lost MB is divided into several smaller subblocks, each pixel in a subblock is directionally interpolated from representative boundary pixels adjacent to the subblock. Experimental results show that the proposed algorithm can recover high as well as low frequency information in lost MB's and provide better visual quality and achieved with 1.5~2.5dB increase in PSNR in comparison with the conventional spatial domain interpolation.

EL-MAC, a contention-based MAC protocol, is proposed in this paper. EL-MAC is efficient and low-latency for the WSN. By cross-layer operating, nodes build the data forwarding chain (DFC) by contending means and staggered wake up in proper time so that packets could be duly forwarded as more hops as possible in a cycle and the latency decline greatly. The simulation results show that EL-MAC has lower latency and higher throughput with less energy consumption on the same environment condition than S-MAC.

The color variations between multi-view video sequences may degrade the inter-view prediction and result in low coding efficiency. In this paper we propose an efficient multi-view video coding scheme using dominant basic color mapping based color correction. The experimental coding results show that color correction has the potential to make multi-view video coding more efficient.

Recent advances in wireless sensor networks have led to many new protocols specifically designed for sensor networks where energy awareness is an essential consideration. Most of the attention, however, has been given to the routing protocols. The paper presents a new clustering scheme named HCMC based on improved Monte-Carlo localization algorithm MCBR. We also exploit multi-hop scheme between clusters to increase the energy efficiency of the protocol. The results show that our approach distributes the energy dissipation more evenly than other clustering schemes, reduces overall energy consumption and improves network lifetime.

In this paper, we present a method for directly computing the surface gradient of each pixel for 3D vision information computation on the basis of shape from shading. In our method, we estimate the unknown surface gradient of a current pixel, using the known surface gradients of its neighbor pixels, and then through several rounds of boundary tracking, we traverse all points and compute their surface gradients. Our method is faster, since each operation is purely local and direct. Besides, by adjusting a control parameter, i.e. iteration accuracy, it can give more precise depth map for a synthetic image.

This paper proposes EL-MAC, a contention-based medium access control (MAC) protocol, which is efficient and low-latency for the wireless sensor network (WSN). EL-MAC introduces duty-cycle and virtual cluster scheme within the framework of S-MAC to reduce energy consumption and to self-organize network. Besides, Inspired by D-MAC, the scheme of data forwarding chain (DFC) is proposed for reducing the latency in multi-hop transmission. The experiment of simulation shows that EL-MAC has lower latency and higher throughput with comparative energy consumption on different traffic load condition than S-MAC.

IPTV can be a new service performed on the Internet in that network transmission and streaming media technologies are getting mature. In this paper, IPTV system infrastructure of UTStarcom, key technologies deployed, and applications will be discussed and evaluated. The key technologies to achieve IPTV services include 1) codec and compression; 2) streaming media; and 3) broadband networks and access to such networks. The implementation of Media Switch IPTV system in Harbin CNC city network is also discussed.

The band structure, optical and defects properties of Ba_{2}TeO are systematically investigated using density functional theory with a view to understanding its potential as an optoelectronic or trans- parent conducting material. Ba_{2}TeO crystallizes with tetragonal structure (space group P4/nmm) and with a 2.93 eV optical band gap 1 . We find relatively modest band masses for both electrons and holes suggesting applications. Optical properties show a infrared-red absorption when doped. This could potentially be useful for combining wavelength filtering and transparent conducting functions. Furthermore, our defect calculations show that Ba_{2}TeO is intrinsically p-type conducting under Ba-poor condition. However, the spontaneous formation of the donor defects may constrain the p-type transport properties and would need to be addressed to enable applications.