Search

Quantum cryptography is a major ingredient of the future quantum internet that promises various secure communication tasks. Quantum conference key agreement (CKA) is an important cryptographic primitive of quantum cryptography, which provides the conference key among multiple users simultaneously. However, quantum CKA is currently far from practical application due to the low conference key rate. Here, we propose a quantum CKA protocol of three users with information-theoretic security. Our protocol only requires phase-randomized weak coherent sources and threshold single-photon detectors, and is anticipated to be experimentally demonstrated over 600 km under current technology. Our scheme can be widely implemented in the approaching large-scale quantum network., Comment: 6 pages, 3 figures, accepted by IEEE Access

In agriculture, many decisions rely on the experiences of farmers, and these decisions are often difficult to quantify with simple numerical values. However, manpower has gradually been replaced by machines in the progress of agriculture. The farmers change from the laborers to decision makers for expanding planting scale or improving quality. The agricultural Internet of Things (IoT) system has become the trend of new agriculture. The data monitoring and automation control system has helped many farmers. In addition, machine learning is also widely used in the agriculture. The irrigation is the most common. The main purpose of this thesis is to effectively use agricultural IoT systems and machine learning to improve traditional agriculture. This thesis provides an ensemble learning irrigation model based on the agricultural IoT system. The IoT system provides the data collection and data monitoring functions, and also supports the website and mobile applications for the convenience of farmers. The irrigation models are also embedded in the agricultural IoT system.

Inability to carry out cohesive narratives has been identified in children with autism spectrum disorder (ASD). However, deriving cohesion measures is often done using manual labeling or relying on expert-crafted features. In this work, we develop a novel LSTM framework to learn the embedded narrative cohesion representation from data directly. Our lexical coherence representation achieves a promising recognition accuracy of 92% in classifying between typically-developing (TD) and ASD children, as compared to 73% by using conventional coherence measures computed from syntactic, word usage, and latent semantic analysis. We perform additional validity analyses on our proposed representation. By experimentally introducing incoherence in the TD's story-telling narratives through word and sentence-level shuffling, the derived lexical coherence representation from these incoherent TD data samples result in a representation closer to those of ASD data samples.

Image inpainting is a technique to repair the damaged part of images or to remove unwanted artificial post-production effects on them. The image after inpainting should be restored to a state which looks like it has never been altered before. In this paper, we proposed an image inpainting technique with a simple mechanism to divide patches of an image into different priority levels. While most literature works need to inpaint these levels in a restricted order, we implemented a scheduling mechanism which frees up the inpainting process from this rigid repair order, allowing for better performances.

The results of the existing literature reveals that the dynamic system is in vast probability of chaos by using the sine function as an auxiliary function. Studies have shown that DCT with magnification owns stronger chaotic structure abilities than any other ones. Relocating matrix of DTC basis function as auxiliary function, which forms the dynamic system with gray image matrix has characteristics in initial value sensitive, multicycle, complexity of bifurcation phenomenon, positive of approximate Lyapunov exponent and so on. There is found to be similar chaotic attractors, whose shapes and contours are similar to the face image of the same person. These attractors used in image recognition, video retrieval and other fields have advantages like simple algorithm, short recognition time and good recognition effect. It is an admirable example of studying chaos in finite field, we ought to study the mechanism of the chaos and other applications in the fields of data processing as a consequence.

In this paper we make improvement to fisheye camera calibration methods. Chessboard corners are used to calculate the relation between 2D image points and 3D world points. In addition, we perform a weighted joint residual refinement. With our fisheye camera calibration method, we can attain better camera parameters and produce a more accurate fisheye undistortion image. Experiment results shows that our method can correct radial distortion and stretched boundary in fisheye images.

We proposed a 3D Around View Monitoring System that is both accurate and efficient. This system consists of images taken from four fisheye cameras which, through an improved calibration process combined with our self-designed triangle-patterned locating method, will be able to automatically determine the homography matrix. A 3D model can then be generated to complete the 3D AVM system. Not only did we propose a new fisheye calibration method to improve the stitching accuracy of the AVM system but we also simplified the once complex registration process.

A strong classifier model oriented to the financial risk warning of listed company has played an important role in the risk analysis of enterprise finance. Based on the BP_Adaboost composed of BP neural network and Adaboost algorithm, a strong classifier model of enterprise finance is designed and established, and is verified with the actual data. In the first, the evaluation index has the largest correlation to the warning result of enterprise finance has been filtered through the significance analysis and factor analysis of typical data. After that, the data item corresponding to the evaluation index is taken into the strong BP_Adaboost classifier model, in order to compute the classification results. Finally, the classification result of the weak BP classifier is computed and compared with the one of the strong BP_Adaboost classifier model. The experimental results show that the accuracy of strong BP_Adaboost classifier is higher than the one of weak BP classifier, and the strong BP_Adaboost classifier has better stability and higher reliability.

Enterprise in financial trouble is a comprehensive event and the enterprise financial situation can be reflected through the liquidity ratio, earnings per share and net assets per share and cash content per share. Artificial neural network method is used to establish the financial early warning model to find the potential financial crisis at an early age. The experiment results show that BP neural network has good characteristics in financial situation evaluation and provides reliabe model for list company financial early warning.

The purpose of this study is to establish a set of accurate and useful financial early warning model for enterprises. In order to achieve this goal, theoretical analysis is combined with previous studies, on the basis of enterprise own characteristic, enterprise financial early warning index system of level 3 is put forward. BP neural network is used to predict financial risk, which is one kind of pattern recognition method. Samples are selected to train the neural network, and the testing samples are used for testing the trained network to verify the accuracy of the trained network. The empirical results confirm that the proposed enterprise financial early warning model can be quite effective for enterprise financial early warning.

Automatic speech recognition (ASR) is a technology which converts the phrases or words spoken by human into text. As a mature technology, ASR has become an alternative input method on many mobile devices, complementing the other input methods operated by hands. Although the technology has been developed for years, the accuracy and computational complexity of ASR have prohibited ASR from being used as the primary input method on mobile devices. While speaker-dependent ASR (SD-ASR) technologies may be used to improve the recognition accuracy, the user is often reluctant to take a time-consuming training process needed to enable SD-ASR for each device. To overcome these problems, we propose a cloud-assisted speech recognition service and its infrastructural design, called CSR, which utilizes servers in the cloud to accelerate ASR, integrates SD-ASR technologies to improve the accuracy of ASR, and populate SD information to enable SD-ASR on multiple mobile device. We have built a prototype to observe the benefits of the CSR service and the issues in load balance, power consumption, and privacy on the client and the server. We show that the CSR service offers fast responses, good accuracy, high availability and good scalability in serving many concurrent users.