Volume 4 Number 3 December 2014
1 |
A Comparative Analysis on Fingerprint Binarization Techniques Abstract: The robustness of a fingerprint authentication system depend on the quality of the binarized fingerprint image. Since the uniqueness of a fingerprint image is determined by the minutiae points which are extracted from a binarized fingerprint image. A very robust binarization process is therefore essential to get the correct set of minutiae points. Thresholding is an effective tool for binarization. In this paper rough set based method is compared with traditional Otsu’s method for binarizing the fingerprint image. The experiments have been conducted on the fingerprint databases FVC2002 and FVC2004. The quantitative metrics such as Relative foreground Area Error (RAE), F-Measure are used to evaluate the performance of the binarization algorithm. Adaptive local threshold based on the mean is used to construct the reference image for RAE computation. The RAE, F-measure of the rough set based method is improved when compared with the Otsu’s method. |
2 |
Trust Based Resource Selection in Cloud Computing Using Hybrid Algorithm Abstract: Cloud computing is experiencing rapid advancement in academia and industry. This technology offers distributed, virtualized and elastic resources as utilities for end users and can support full recognition of “computing as a utility” in the future. Scheduling distributes resources among parties which simultaneously and asynchronously seek it. Scheduling algorithms are meant for scheduling and they reduce resource starvation ensuring fairness among those using the resources. Most Task-scheduling cloud computing procedures consider task, resource requirements for CPU and memory, and not bandwidth. This study suggests optimizing scheduling with BAT-Harmony search hybrid algorithm. |
3 |
Artificial Neural Networks’ Application in Weather Forecasting – Using RapidMiner Abstract: Weather forecasting is a crucial phenomenon in today’s world. Though weather prediction is completely automated, with the help of tools like Weather Research & Forecasting (WRF), Advanced Research WRF (ARW), Weather Processing System (WPS), it’s a ever challenging and a topic of interest because prediction is not an accurate always. Weather forecasting is a continuous, high dimensional, dynamic and complicated process because it involves many entities of the atmosphere. The parameters required to predict the weather are enormously complex such that there is uncertainty in prediction even for a short period. The property of artificial neural networks is that they not only analyze the historical data, but also learn from it for future predictions make them suitable / ideal for weather forecasting. Weather prediction can be simplified by using the artificial neural networks (ANN) with back propagation for supervised learning using the data collected at a particular station at a specified period. After training the model, they are used to predict the weather conditions. As an experimental method, the model is made known to predict the values as unknown values. The output is promising and motivates us to work more towards this goal. |
4 |
Relevance Feedback Mechanism for SMS based Literature Retrieval in Indic Languages Abstract: The concept and realization of ‘Information Pulling’ on handheld mobile devices facilitated an easy and effective information access. The researchers and developers are applying their efforts to concur mobile handsets into a timely business processing and information access terminals. Many of these applications use fixed format query answer method. This popularly used application type sends information pushing messages to mobile subscribers related to a specific domain. Agricultural messages, banking messages, health care messages, astronomy messages and advertising messages are examples of these types. This type of system is categorized as “Service Initiated Communication” system. Another category applies an information retrieval methodology for pulling information on mobile handset from the service server as per user’s demand. This type of system is known as “User Initiated Communication” system. In this case the demand of information is either in fixed form or flexible form queries. Flexible form SMS query in natural language format for information access can be considered as recent research domain in this regard. We have developed this second type of SMS based information system for Indic Language Literature. The system applies Vector Space Model for a suitable knowledge representation and an appropriate query-document similarity mapping scheme. In addition, we have focused on the development of a relevance feedback mechanism in order to improve the relevance of the responses of our system. Storage structure of VSM is tailored to represent the specified feature of Self Tagging Indic Literature Documents. The informative tags in documents, content, terms and the respective location of the terms are stored in the document’s term vector. This paper elaborates this modified VSM structure and the relevance improvement mechanism in detail. The results are analyzed and discussed using Discounted Cumulative Gain. |
5 |
Ear Biometrics for Automatic Index Segmentation Using Canny Edge Detection Abstract: A biometric system is essentially a pattern recognition system which uses a Specific physiological or behavioral characteristic of a person to determine their identity or verify a claimed identity. An Ear biometrics for automatic index segmentation algorithm for grayscale image is proposed by applying canny edge detection, distance metric, Euclidian distance and Pythagorean Theorem. In this paper, we have performed the identical to ear image, comparison of the different applied models being currently used for the ear image modeling, processing the details of the algorithms, methods and finally tracking the error and limitation from the input database for ear identification. |
6 |
Analysis of Dimensionality Reduction in Intrusion Detection Abstract: Intrusion detection system is an important technology in the market sector as well as in the area of research. Intrusion detection is considered a useful security tool that assists in preventing attacker’s access to networks or systems. The determination of genuineness of packets is a key issue and various approaches of classification have been presented. The complexity of a classifier is greatly reduced if the numbers of attributes in a data set are reduced. Analysis of dimensionality reduction and it is impacting thereof is the objective of our study. An experimental study is carried out to build up a classifier on a standard dataset of network traffic data that includes normal packets and abnormal packets. A rough set theory and information gain approaches are employed to reduce dimensionality of network traffic data set. The features obtained by the rough set theory and information gain are used to train and test the J48 classifier. A comparative analysis of the results obtained a reduced attribute set and original attributes are presented. The results shows that the performance of J48 classifier with the reduced attributes (rough set and information gain) is better, which is at the cost of time. |
7 |
Comparative Analysis of Saptial Filtering Techniques in Ultrasound Images Abstract: This paper gives the knowledge about the various filter techniques applied in the speckle noise removal from ultrasound fetal images. In many despeckling filters available in speckle reduction, some are best suited in ultrasound speckle noise images. The despeckle image evaluation quality measurements RMSE AND PSNR are compared to the ultrasound images in despeckling for the spatial domain filter. |
8 |
An Empirical Analysis of Flame and Fuzzy C-Means Clustering for Protein Sequences Abstract: Biological data have to be analyzed, interpreted and processed to deal with the problems in life sciences. Bioinformatics addresses the biological problems using computational methods. Clustering is one of the computational techniques for analyzing biological data. Clustering protein sequences into families with similar patterns is important in Bioinformatics. Many clustering algorithms are available for rapid development of protein sequences. In this paper, we compare and evaluate the performance of two clustering algorithms, namely fuzzy c-means and flame for protein sequences. First, we describe each clustering method and compare them through the validity indices and execution time as well. |
9 |
Elliptic Curve Cryptography based key generation from the fusion of ECG and Fingerprint Abstract: This article deals with the new innovative model for cryptographic key generation from the fusion of Electrocardiogram (ECG) and fingerprint using Elliptic Curve Cryptography (ECC). Among all the biometrics, ECG and Fingerprint is used to generate the key since ECG provides intrinsic liveliness detection and fingerprint based identification is highly scalable. After preprocessing the biometric traits, the features are extracted. Then the extracted features are fused to generate the cryptographic key. ECC is used as many mathematicians proved that elliptic curve gives the best solution for cryptography. The generated cryptographic key using the proposed method is smaller when compared with RSA. |
10 |
Rough Set Theory Approach to Generating Classification Rules Abstract: Data mining is the process of extracting hidden information from large databases. Rule induction is the most common data mining technique. Dimensionality reduction gives an optimal subset according to an objective function. Rule generation is one of the important processes in the knowledge discovery system. Rough set approach of generating rules can be used to increase the correct predictions by identifying and removing redundant variables. In this paper we emphasize the role of Reducts, Core and their approximations. Data from UCI repository have been taken to exhibit rules for soybean data set by using ROSETTA. |