The Modares Journal of Electrical Engineering
Year:2014 | Volume:14 | Issue:3|Papers Count:6

in the era of information, data which are worthwhile asset of human, organizations and enterprises have become such sophisticated that the conventional approaches and methods are not usable anymore, or not efficient at least. Such complexity which is known as the Big Data problem is the affordable extraction of value from big data sets that we are encountered in many recent applications e.g., e-business, scientific research, monitoring, search engines, social networking, etc.. Big Data complexities are instantiated by three major dimensions, high Volume, high Variety, and high Velocity (a.k.a.3Vs). The first and most essential step in data management (also for Big Data management) is designing and employing a proper data model, as the footstone of the other data management activities such as R& D of DB languages, DBMSs, tools, methods, algorithms, etc.. In this paper, a proper data model for Big Data is designed and proposed in which the properties required for Big Data problem (i.e., to be integrated, complete, scalable, flexible, compatible, and efficient) are considered. As a data model, data representation is designed and implicit integrity constraints are presented for the proposed HNG (Hyper Nested Graph) data model. Experimental evaluation results show that the proposed data model outperforms other currently used data models such as the document-based, graph document-based, and graph- based data models in terms of response time.

This paper introduces a method for abnormal event detection in video. The video is divided into a set of cubic patches. A new descriptor for representing the video patches is proposed. This descriptor is created based on the structure similarity between a patch and nine neighboring patches of it. All training normal patches in respect to the proposed descriptor are represented and then modeled using a Gaussian distribution as the reference model. In test phase, those patches which are not fitted to the reference model are labeled as anomaly. We have evaluated the proposed method on two UCSD1 and UMN2 popular standard benchmarks. The performance of the presented method is similar to state-of-the-art methods and also is very fast.

Performance of automatic speech recognition (ASR) systems degrades in noisy conditions due to mismatch between training and test environments. Many methods have been proposed for reducing this mismatch in ASR systems. In recent years, deep neural networks (DNNs) have been widely used in ASR systems and also robust speech recognition and feature extraction. In this paper, we propose to use deep belief network (DBN) as a post-processing method for de-noising Mel frequency cepstral coefficients (MFCCs). In addition, we use deep belief network for extracting tandem features (posterior probability of phones occurrence) from de-noised MFCCs (obtained from previous stage) to obtain more robust and discriminative features. The final robust feature vector consists of de-noised MFCCs concatenated to mentioned tandem features. Evaluation results on Aurora2 database show that the proposed feature vector performs better than similar and conventional techniques, where it increases recognition accuracy in average by 28% in comparison to MFCCs.

In this paper, a novel fuzzy connectionist system for incremental online learning and knowledge discovery called Population-based Automatic Fuzzy Neural Network (PAFuNN) is demonstrated in detail. PAFuNNs evolve out of incremental learning. New connections and neurons are created based on a population of samples while operating the system which has the advantage of controlling the number of neurons involved and leads to the low complexity of the network. Learning Automata is implemented in order to optimize the network parameters including sensitivity and error thresholds to enhance the performance of the entire system. Afterward, the proposed method is compared with Evolving Fuzzy Neural Network (EFuNN) as a general online learning machine on two case study datasets consisting of gas furnace and iris data for prediction and classification tasks leading to the thorough analysis of the effects of selecting appropriate automata. Less complex, more accurate and robust results are obtained for the proposed method in comparison with the EFuNN.

In this paper, we proposed a data-hiding scheme based on Run length matrix. In a previously proposed method, a technique based on texture classification was introduced where four statically features extracted from run length matrix; then best cover images are selected based on these features. Using appropriate features for comparing images from undetectability viewpoint, guarantees, less detectability of stego images and consequently, enhances security of the steganography algorithms. Based on this idea, in this paper, more features are extracted from run length matrix to select the best covers. Our method is examined with feature based and wavelet based steganalysis algorithms. The results illustrate the effectiveness and benefits of the proposed method.

In recent years using wireless sensor networks (WSNs) in applications, such as disaster management and security surveillance have been increased. A lot of sensors in these applications are expected to be remotely deployed in unattended environments autonomously. To support scalability, nodes are often grouped into disjointed and mostly non-overlapping clusters. Every cluster has a leader that is known as a cluster-head (CH). The CH may be selected by the sensors in the network or pre-assigned by the network designer. These networks require effective communication protocols to be energy efficient and increase network quality. In this paper, a self-organization routing protocol for wireless sensor networks is presented by using hierarchical protocols and considering the position of CHs regarding to each other which is called "Probabilistic Selection of Cluster-head based on the Nearest possible Distance of Cluster-head". In addition to increase network lifetime, it causes to increase scalability of the network, optimal use of communication bandwidth and improve some of qualitative parameters of the sensor networks. Proposed method has little overhead control and can find appropriate CHs with local information. In this paper, simulation is done by the NS-2 simulator, and simulation results show this protocol could lead to increase environment monitoring, improve network lifetime, throughput and some qualitative sensor network parameters by improving the clustering process of all the routing protocol. WSNs that aren’t considered CHs distribution (LEACH protocol here).