Advanced Materials Research Vols. 756-759

Abstract: With the rapid developing of wireless communication technology and the growing user service, the demand on the system capacity from users is growing. In the LTE network, the uneven load distribution makes that network resource is not fully utilized, which results that the capacity is limited. In this paper, a novel optimizing mechanism of network load balancing based on the network manager is proposed. The eNBs in the network will broadcast the current load status by the way of adapting a new message format with the life cycle bit when the load of the cell is changed, then the serving eNB can decide which part of users in the cell should transfer to the neighboring cells and send a request to the network manager. The network manager will calculate whether the transferred users make the neighboring cell load exceed the threshold. If not, the users can be transferred to the neighboring cell and release the serving cells load burdon. Otherwise, the serving cell will search other neighboring cells to transfer the users. The simulation results show that the proposed mechanism will not consume more network resource and improve the network throughput significantly.

Abstract: Analyze a human visual model based wavelet domain digital watermark algorithm and improved it. First conduct multi-level wavelet decomposition on original image and modify medium frequency coefficient, and then adapt the algorithm to embed the watermark into the source image. The experiment shows that the watermark embedded image enjoys better imperceptibility and robustness to JPEG compression, add gaussian noise, contrast enhancement e.t. traditional image processing operations.

Abstract: In this paper, we propose a novel medical color image digital watermarking algorithm, which can realize watermark capacity, transparency and robustness effective compromise. We encrypt the watermark image by Arnold Transform at first and then take it into a random sequence. The principal component of medical image is divided into some sub-blocks with 4×4 and each block is transformed by DCT. One bit of watermark is embedded by comparing the size of specific DCT coefficients in the adjacent sub-blocks. Watermark can be extracted without the original medical image. The experimental results show that the digital watermarking algorithm has a high transparency and big capacity as well as certain robustness.

Abstract: The aliasing due to subsampling and the blur from the finite detector size can decrease quality of the images and make fine details and structures difficult to interpret. High resolution images can be reconstructed from several adjacent frames in a sequence by reconstruction process. This paper describes the high resolution image reconstruction method based on wavelet domain. In this approach both the image sequences and the degradation operator are presented by orthogonal wavelet with compact support. Experimental results demonstrate that the proposed method is effective to improve image details.

Abstract: The image is a kind of important information source, through the image processing can help people understand the connotation of the information. However, the image in the process of generation and transmission by all kinds of the noise, the information processing, transfer and storage caused tremendous influence. So the image denoising always all is the computer image processing and computer in the vision of a research focus. Proposed an algorithm of image denoising based on the local structural similarity. Which utilizes the redundant information, and by establishing a similar function to the search area calculation of point and to pixels similarity of weights, and then to the search area at the weighted, obtained the last to pixels gray value varies. This algorithm in texture, of the edge information denoising ways than the current many denoising algorithm are excellent.

Abstract: In order to solve the problem of how to improve the scalability of data processing capabilities and the data availability which encountered by data mining techniques for Data-intensive computing, a new method of tree learning is presented in this paper. By introducing the MapReduce, the tree learning method based on SPRINT can obtain a well scalability when address large datasets. Moreover, we define the process of split point as a series of distributed computations, which is implemented with the MapReduce model respectively. And a new data structure called class distribution table is introduced to assist the calculation of histogram. Experiments and results analysis shows that the algorithm has strong processing capabilities of data mining for data-intensive computing environments.

Abstract: We review a recent neural implementation of Canonical Correlation Analysis and show, using ideas suggested by Ridge Regression, how to make the algorithm robust. The network is shown to operate on data sets which exhibit multicollinearity. We develop a second model which not only performs as well on multicollinear data but also on general data sets. This model allows us to vary a single parameter so that the network is capable of performing Partial Least Squares regression to Canonical Correlation Analysis and every intermediate operation between the two. On multicollinear data, the parameter setting is shown to be important but on more general data no particular parameter setting is required. Finally, we develop a second penalty term which acts on such data as a smoother in that the resulting weight vectors are much smoother and more interpretable than the weights without the robustification term. We illustrate our algorithms on both artificial and real data.

Abstract: We propose a new self-organizing neural model that performs principal components analysis. It is also related to the adaptive subspace self-organizing map (ASSOM) network, but its training equations are simpler. Experimental results are reported, which show that the new model has better performance than the ASSOM network.

Abstract: How to improve the frequency resolution, especially in the low frequency bands, is an important problem for many applications. A novel approach based on sparse representation was proposed for this problem. In this work, we first discussed the design of the over-completed dictionary with the high frequency resolution, and then Matching Pursuit was advised to perform sparse decomposition using Matching Pursuit Tool Kit. Next, the frequency information was extracted from the code book of the results of Matching Pursuit. Finally, the experiments demonstrated the improvement of the frequency resolution for the proposed approach.

Abstract: Existing data clustering method lacks considering of latent similar information existing among words,and it leads to unsatisfactory clustering result.Aiming at Chinese short message text clustering,this paper proposes a clustering algorithm based on semantic.It offers Chinese concept,and the measuring methods to calculate the similarity degree about words and Chinese short message text.It completes the clustering of Chinese short messages text through fission downwards and mergence of twos upwards.Experimental results show that this algorithm has better clustering quality than traditional algorithm.