AASRI Procedia最新文献

In this work, an adaptive algorithm with variable tap-length is proposed for diffusion networks, where all nodes exchange information with their connected group without a fusion center and utilize both exchanged information and their local observation to make their individual estimations. Based on a fractional tap-length solution and the fully distributed structure of networks, we propose the algorithm to resolve the estimation problems on both tap-length and tap-weights. The proposed algorithm with adaptive tap-length structure enables the networks to reduce communication energy and network resources. The convergence properties of this algorithm are confirmed by the simulation results.

Spam and spam filters are contrarious components of a complex interdependent social ecosystem. Traditional spam filtering techniques or systems are usually designed and deployed individually that neglect the distributed and bulk characteristics of spam. This paper proposes a reputation-based collaborative anti-spam approach. This approach, adopting fingerprinting technique, evaluating reporters’ trust, achieved better performance and robustness than the state-of-the-art in comparison experiments on several known email corpora.

This paper first describes the relationship between the smart grid and smart AVC. By some problems of the conventional AVC's practical application connecting with the actual situation of the current grid development, this paper points out that the traditional AVC is unable to adapt to the development needs of the current grid and the tendency of intelligent AVC development.Through the research and the comprehensive exposition of the embedded Smart AVC, including intelligent AVC's new features, it points out that the future development direction of the Smart AVC and some features needs.

This paper proposes a robust, efficient and scalable distributed Arabic handwriting OCR system based on a parallel FastDTW algorithm via cloud computing technologies. The three techniques Hadoop, MapReduce and Cascading are used to implement the parallel FastDTW algorithm. The experiments were deployed on Amazon EC2 Elastic Map Reduce and Amazon Simple Storage Service (S3) using a large scaled dataset built from the IFN/ENIT database.

Traditional method to learn Chinese calligraphy needs long hour's practice and the practice process often makes learners feel boring. And learners cannot get instant evaluation result during training. In this paper we introduce a new mobile Chinese Calligraphic training system based on virtual reality. The system is inspired by the way of tradition way to learn calligraphy which includes three main steps: a) learning stroke order of character; b) copying writing; c) real writing. Firstly the system demonstrates the right stroke order of calligraphic character with 2D and 3D mode. Secondly learner can write on iPad screen to practice copying writing. Meanwhile the leaner's writing track will be captured and evaluated according the similarity. The result of user studies shows that our system can successfully guide learners to improve their writing skills and motive them to learn calligraphy.

Focusing on the existing location technology limited the performance of the algorithm, distributed data location strategies DLPSL based on weights of jump table is presented to solve the efficiency problem. Weights are added to nodes of jump table, so that the high rate of location of the storage node priority is found to short the search path, and improve the storage and location efficiency. System analysis shows that, the node's insertion, deleting, and location on DLPSL are more efficient than single-linked list storage structure and skip graphic. Its time complexity is 0(log n), and space complexity is 0(n).

When pervasive computing is used to provide technology-driven assistive healthcare, there is a need for the system to be as sophisticated and adaptable as possible, while also being as transparent as possible for both subject and caregivers. A personalized, extensible hybrid reasoning framework has been implemented for the CARA (Context Aware Real-time Assistant) system which aims to satisfy these design goals. It provides context-aware sensor data fusion (including medical and environmental sensors) and incorporates anomaly detection mechanisms that support Activity of Daily Living (ADL) analysis and alert generation. The hybrid reasoning architecture incorporates both rule-based and case-based reasoning; this enables CARA to be more robust and to adapt to a changing environment by continuously retraining with new cases. The rules used for anomaly detection in a smart-home situation are given in a structured natural language, allowing subject or caregiver to inspect and, if appropriate, modify these rules; this supports the goal of transparency. For the case-based reasoning part, attention is drawn to the transparency issues that arise in the evaluation criteria used and interpretation of results.

The accuracy of the interpolation method for identification of nonlinear dynamical systems based on the Volterra model in the frequency domain is studied. To highlight the n-th partial component in the response of the system to the test signal the n-th partial derivative of the response using the test signal amplitude is found and its value is taken at zero. The polyharmonic signals are used as test ones. The algorithmic and software toolkit is developed for identification processes. This toolkit is used to construct the informational model of test system. The model is built as a first, second and third order amplitude–frequency characteristics and phase–frequency characteristics. The comparison of obtained characteristics with standard is given.

In this paper, we present a new method for analysis of cardiac sound signals containing murmurs using constrained tunable-Q wavelet transform (TQWT). The fundamental heart sounds (FHS) and murmurs are separately reconstructed by suitably constraining TQWT. The segmentation of reconstructed murmurs into heart beat cycles is achieved using cardiac sound characteristic wave-form (CSCW) of reconstructed FHS. The frequency domain based approximate entropy, spectral entropy, Lempel-Ziv complexity, and time domain Shannon entropy are computed for each segmented heart beat cycles for least squares support vector machine (LS-SVM) based classification. The experimental results are included to show the effectiveness of the proposed method.

In robotics, pose errors are known as positional and rotational errors of a given mechanical system. Those errors are commonly produced by the play among joined components, commonly known as joint clearances. Predicting pose errors can be done via the formulation of two optimization models holding continuous domains, which belong to the NP-Hard class of problems. This paper focuses on providing rigorous and reliable solution to this problem by using constraint programing.