Advanced Materials Research Vols. 734-737

Abstract: In this paper, we present an approach of three-dimensional human face pose correction with the normal vector alignment algorithm. We detect three feature points on a human face through calculating discrete Gaussian curvature. Then we calculate the three feature points plane of the normal direction. The face pose is corrected from the normal vector direction. This method is small amount of calculation and wide applicability. The experimental results show that the correction effect is good.

Abstract: To release warning information correctly and timely, the team tried to design a software called Chongqing Emergency Warning Information Release Software (CEWIRS) based on the analysis of the needs of Chongqing emergency warning information releasing work, and developed some modules. This software proves to be great help to improve efficiency of Chongqing emergency warning information releasing work. After weighing up the costs and benefits of developing CEWIRS, the team suggests that such means as website and RSS should be paid more attention to get warning information received by more people at lower cost.

Abstract: With the widely use of smart phone in China, all inputs and routes packets streams to the Content Distribution Service (CDS) switching centers. Each produces up to 1.5 terabytes arriving every day. Normally, the job of the switch is to transmit data. Obviously, the ordinary database cannot handle the massive dataset and complex ad-hoc query. In this paper, we propose DeepMR, a MapReduce deep service analysis system based on Hive/Hadoop frameworks. A distributed file system HDFS is used in DeepMR for fast data sharing and query. DeepMR also optimizes scheduling for switch analysis jobs and supports fault tolerance for the entire workflow. Our results show that the model achieves a higher efficiency.

Abstract: In this paper, a matching theorem for weakly transfer compactly open valued mappings is established in GFC-spaces. As applications, a fixed point theorem, a minimax inequality and a saddle point theorem are obtained in GFC-spaces. Our results unify, improve and generalize some known results in recent reference.

Abstract: Monitoring modulation type of the detected signal is the most important intermediate step between signal detection and demodulation. The back propagation neural network (BPNN) was widely used in constructing modulated signal classifier in the field of automatic modulation classification (AMC). There are many visible features in the back propagation (BP) algorithm including adaptive learning, the ability of fault tolerant, etc. However, this algorithm has two main disadvantages, such as the slow convergence speed and easily falling into the local minimum. This paper presents a novel modulation classifier using BPNN trained with swarm intelligence algorithms (SIA), for the sake of overcoming these deficiencies. The initial weights and thresholds of BP neural network were optimized by SIA. As the SIA has an excellent global search property, this classifier can consume less training time and improve the automatic modulation type identification rate.

Abstract: Considering that the particle swarm optimization (PSO) algorithm has a tendency to get stuck at the local solutions, an improved PSO algorithm is proposed in this paper to solve constrained optimization problems. In this algorithm, the initial particle population is generated using good point set method such that the initial particles are uniformly distributed in the optimization domain. Then, during the optimization process, the particle population is divided into two sub-populations including feasible sub-population and infeasible sub-population. Finally, different crossover operations and mutation operations are applied for updating the particles in each of the two sub-populations. The effectiveness of the improved PSO algorithm is demonstrated on three benchmark functions.

Abstract: This paper presents a new type of deaf-mute sign language recognition system by combining mobile communication platform and the mobile communication terminal equipment. The system can implement the vision-based sign language recognition and translation. A Standard Sign Language Database is established in this system. Multi-national and multi-language sign language recognition can be completed by the following training using the database. In order to improve the accuracy of the recognition of similar sign language, an improved HMM sign language recognition method is used in this paper. The angle information of sign language which can be achieved by the traditional data-glove is introduced in the system on the basis of visual methods, makes the system taking into account these two recognition technology. The system can be implemented in ordinary mobile terminal equipment. Low cost and popularity of sign language recognition device can be realized. The deaf-mutes communicate for barrier-free anytime and anywhere by the application.

Abstract: Semantic knowledge-base has important meaning for increasing the deepness of NLP. Some comparatively mature Semantic knowledge-base such as WordNet, HowNet and Thesaurus was developed by manpower, and has many difficulties on actual application. In order to capture Chinese word knowledge of relating status moue automatically and demonstrably, this paper presented the concept of word correlation and a calculation method of word correlation based on statistic. Then a correlation net based on Chinese words which have strong domain characteristic was built. In order to resolve the difficulty of processing the huge amount of data, a hard disk storing method of array segmentation was designed. The semantic knowledge gained by the experiment had the advantage of empiricism. It is veracity and generalization is strong so it can play an important role in many fields such as text categorization, text retrieval, text filtering, etc.

Abstract: To characteristics the near infrared spectrum of corn, we proposed a preprocessing algorithm combining multiplicative scatter correction (MSC) and direct orthogonal signal correction (DOSC). In this article, we compared the SG first derivative correction, MSC, DOSC and the new algorithm. And using partial least squares analysis (PLS) built the quantitative analysis model. The results show that the preprocessing algorithm combining with the MSC and DOSC can improve the prediction accuracy of all components. This work has some practical and scientific value.

Abstract: TFT-LCD panel defect detection has been one of the difficulties in this field because of fuzzy defect boundary, low contrast between defects and background, and low detection speed. The structure of TFT-LCD panels and classification are introduced. Through the analysis of panel defect features, current detection methods for the TFT-LCD panel defects are reviewed. The key technologies of feature extraction and defect classification are analyzed in the defect image recognition of TFT-LCD panel. Meanwhile the methods of fuzzy boundary defect segmentation, image subtraction and image filtering are also discussed. Finally, the characteristics and advantages of these detection methods are concluded, and several key issues for the TFT-LCD defect detection have been proposed for future development.