Informatics Publishing Limited

J. Jebathangam ¹, S. Purushothaman ², P. Rajeswari ¹

Affiliations
1 Mother Teresa Women’s University, Kodaikanal, IN
2 Institute of Technology, Haramaya University, ET

Abstract

This paper presents a combination of wavelet with echo state neural network in identifying the microcalcification (MC) in a mammogram image. Mammogram image is decomposed using Daubauchi wavelet to 5 levels. Statistical features are extracted from the wavelet coefficients. Training of the ESNN/BPA is done using the features as inputs to the network along with a labeling of presence or non-presence of MC. The classification performance of ESNN is compared with back propagation algorithm.

Keywords

Microcalcification, Wavelet, Neural Network, Echo State Neural Network, Backpropagation Algorithm.

Full Text

     

P. Sujatha ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 School of Computing Sciences, Vels University, Chennai, IN
2 Udaya School of Engineering, Kanyakumari – 629 204, Tamil Nadu, IN
3 Mother Therasa University, Kodaikanal, IN

Abstract

The covert communication based on steganography is a challenging technology for governments. Using this most powerful technique terrorists and spies communicate with each other to exchange their plan which is not detected by law enforcement. In order to avoid the misusage of steganographic technique, the government needs to find out some powerful techniques to detect the existence of the hidden data in the digital media. This leads to the concept of steganalysis that is used in many fields such as digital forensics, medical imaging, and journalism. Apart from all modern sciences and technologies, Artificial Neural Network (ANN) plays a vital role in capturing and representing both linear and non-linear relationships. ANN is an intelligent system which enables machines to solve problems like human by extracting and storing the knowledge. Hence to incorporate intelligent method for steganalysis, this paper implements Artificial Neural Network to overcome the drawbacks of the conventional methods. The most powerful Radial basis function algorithm is proposed in this paper since it is more suitable for non-linear data. This paper concentrates on detecting the hidden information for digital forensics application.

Keywords

Artificial Neural Network, Covert Communication, Radial Basis Function (RBF), Steganography, Steganalysis.

Full Text

     

V. Savithri ¹, S. Purushothaman ²

Affiliations
1 Mother Teresa Women's University, Kodaikanal, IN
2 Udaya School of Engineering, Kanyakumari District-629204, IN

Abstract

The objective of this work is to develop and implement an investigation carried out on carotid artery. The proposed method categorizes the carotid artery subjects into normal and diseased subjects. Ultrasound image videos of the artery are used as the data. The frames of the video are processed to know the properties of the artery. In order to extract properties, image processing techniques have been used on each frame. The features extracted from the frames are consolidating to know the conditions of the artery. These features are accurately analysed to know the status of the artery by using intelligent techniques like artificial neural network and fuzzy logic. In artificial neural network using back propagation algorithm and by fuzzy logic system provides higher classification efficiency with minimum training and testing time. It helps in developing medical decision system for ultrasound artery images. It can also be used as secondary observer in clinical decision making.

Keywords

Artery, Boundary Detection, Intima Media Thickness, Neural Network, Ultrasonic.

Full Text

     

A. Muthukumaravel ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 VELS University, Chennai-600117, IN
2 Udaya School of Engineering, Kanyakumari-629204, IN
3 Mother Teresa Women's University, 624101, IN

Abstract

Concurrency control is the action of synchronizing operations issued by concurrently executing programs on a shared database. Concurrency control (CC) is the key concept for proper transactions on objects to avoid any loss of data or to ensure proper updating of data in the database. This paper presents the development of Fuzzy Logic for concurrency management while modeling Bolted connection drawing using Autodesk inventor 2008. While implementing concurrency control, this work ensures that adjacent shapes in the drawing cannot be accessed by more than one person. The Fuzzy Logic learns the objects and the nature of transactions to be done based a threshold value. Learning stops once all the objects are exposed to Fuzzy Logic. We have attempted to use Fuzzy Logic for storing lock information when multi users are working on same Bolted connection drawing.

Keywords

Concurrency Control, Fuzzy Logic, Bolted Connection, Transaction Locks, Computer Aided Design Database.

Full Text

     

Shoba Rani ¹, S. Purushothaman ²

Affiliations
1 Mother Therasa University, Kodaikanal, IN
2 Sun College of Engineering & Technology, Kanyakumari – 629902, Tamil Nadu, IN

Abstract

The work focuses on segmentation of various textures of a given image. In addition, retrieval of the desired component of the image is done. The segmentation and retrieval is based on the texture features of the image. The goal of the work is to develop a promising technique to segment the textures of the image. Basically,we use a set of GABOR filters segmenting the given image and using artificial neural network (ANN) for extracting the relevant information. 

The theme of the work is based on GABOR filters which are based on the famous Gaussian function. Using Gabor filters, does the extractions of features of the various textures in the given image. By using the obtained features, subsequent labeling is done by (Backpropagation algorithm) ANN. The nature of the work involves taking an image texture as input and getting the partitioned or segmented textures as output. It can be very well justified in doing this work as automatic segmentation, identification and classification and as an important role image processing. The future of the work is unlimited.

Keywords

Texture Image Segmentation, Image Retrieval, Gabor Filters, Artificial Neural Network.

Full Text

     

T. Suganthi ¹, S. Purushothaman ²

Affiliations
1 Mother Teresa Women’s University, Kodaikanal, Tamilnadu, IN
2 Sun College of Engineering & Technology, Kanyakumari – 629902, Tamil Nadu, IN

Abstract

Satellite images are used to understand the land information. But it is difficult to identify the area which is below the clouds. So it is very important task to detect and removal the cloud images. The aim is to give the method for automatic detection and removal of cloud and its shadow contamination from satellite images. After detecting the contamination the method is used to replace the data from different images of the same area to minimize the cloud contamination effect. The Adaptive Brightness Threshold Algorithm (ABT) and Counter Propagation Network (CPN) are used for detection. CPN is able to show the better performance than ABT.

Keywords

Cloud Detection, Threshold, Counter Propagation Network, Satellite Images.

Full Text

     

Shoba Rani ¹, S. Purushothaman ²

Affiliations
1 Department of Computer Science, Mother Therasa University, Kodaikanal, IN
2 Sun College of Engineering and Technology, Nagercoil, IN

Abstract

Texture is a ubiquitous experience and can describe a variety of natural phenomena with repetition, such as sound (background noise in a machine room), motion (animal running), visual appearance (surface color and geometry) and human activities (daily lives). Since reproducing the realism of the physical world is a major goal for computer graphics, textures are important for rendering synthetic images and animations. However, as textures are so diverse it is difficult to describe and reproduce them under a common framework. In this paper, new methods for synthesizing textures are presented. The initial part of the paper is concerned with a basic algorithm for reproducing image textures. The limitations of traditional methods can be overcome by the proposed approach based on multi-resolution (search neighborhoods and tree-structured vector quantization) analysis. The paper concerns with various extensions of the basic algorithm; the extensions concentrate on either reproducing textures of different physical phenomena such as motions, or creating textures in novel ways in addition to mimic existing ones.

Keywords

Texture, Wavelet, Multiresolution Analysis and Image Retrieval.

Full Text

     

R. Bhuvana ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 Department of Computer Science, Vels University, Chennai, IN
2 PET Engineering College, Vallioor-627117, IN
3 Mother Teresa Womens University, Kodaikanal, IN

Abstract

Depression is a serious and widespread public health challenge. This paper propose neural network algorithm for faster learning of psychological depression data. Implementation of neural networks methods for depression data mining using back propagation algorithm (BPA) and Echo state neural network (ESNN) are presented. Experimental data were collected with 21 input variables and one output for working with artificial neural network (ANN). Using the data collected, the training patterns and test patterns are obtained. The input patterns are pre-processed and presented to the input layer of ANN In order to find the optimum number of nodes required in the hidden layer of an ANN, a method has been proposed, based on the change in the mean squared error dynamically, during the successive sets of iterations. The output of BPA is given as input to ESNN. The network trained with transformed vectors is seen to require the least computational effort. The work proves to be an efficient system for diagnosis of depression.

Keywords

Hamilton Rating Scale (HRS) Depression Data, BPA, ESNN.

Full Text

     

P. Karthikeyan ¹, S. Purushothaman ²

Affiliations
1 Department of Civil Engineering, CMJ University, IN
2 PET Engineering College, Vallioor-627117, IN

Abstract

In regions of high seismicity, infilled frames are commonly used for low and medium-height buildings. "Infilled frame" is a composite structure. It is formed by one or more infill panels surrounded by a frame. Infilled frame also refers to the situation in which the frame is built first and then infilled with one or more masonry panels. The primary function of masonry was either to protect the inside of the structure from the environment or to divide inside spaces. The presence of masonry infills helps the overall behavior of structures when applying lateral forces. The lateral stiffness and the lateral load capacity of the structure largely increase when masonry infills are considered to interact with their surrounding frames. In this paper, ANSYS 14 software is used for analyzing the infill frames. Echo state neural network (ESNN) has been used to supplement the estimation of stress values of the proposed infill frame model. The number of nodes or reservoirs in the hidden layer for ESNN algorithm varies depends upon the accuracy of estimation required. Exact number of reservoirs is fixed based on the trial and error method, through which the accuracy of estimation by the ESNN is achieved.

Keywords

Echo State Neural Network (ESNN), Reservoir, Processing Elements (PE), Finite Element Method (FEM), Equivalent Stress.

Full Text

     

B. Shaji ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 VELS University, Pallavaram, Chennai-600117, IN
2 PET Engineering College, Vallioor-627117, IN
3 Mother Teresa Women's University, Kodaikanal-624102, IN

Abstract

Wavelet decomposition has been applied to mammogram image to obtain four different coefficients for 5 levels of decompositions. The coefficients are: low frequency coefficients (A), vertical high frequency coefficients (V), horizontal high frequency coefficients (H), and diagonal high frequency coefficients (D). The features of the mammography image are obtained using the wavelet transform selecting the different levels of decompositions. The proposed method presents a new classification approach to microcalcification (MC) detection in mammograms using wavelet and back propagation algorithm (BPA) Neural Network. These features obtained from wavelet are representation of MC as well as other information of the image. Daubauchi wavelet has been used to decompose image to 5 levels. Statistical features are extracted from the wavelet coefficients. Training the BPA with features and testing the BPA to identify the presence of MC has been done. The percentage identification is above 96.2%. The performance of the proposed method based on the quality of the mammogram image.

Keywords

Mammogram, Microcalcification, Wavelet Transform, Back Propagation Neural Network.

Full Text

     

Y. Kalpana ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 VELS University, Pallavaram, Chennai-600117, IN
2 PET Engineering College, Vallioor-627117, IN
3 Mother Teresa Women's University, Kodaikanal-624102, IN

Abstract

Intrusion detection is the art of detecting computer abuse and any attempt to break into networks. As a field of research, it must continuously change and evolve to keep up with new types of attacks or adversaries and the ever-changing environment of the Internet. To make networks more secure, intrusion-detection systems (IDS) aims to recognize attacks. Artificial neural networks (ANN) based IDS were implemented and tested. The goal for using ANNs for intrusion detection is to generalize from incomplete data and able to classify data as being normal or intrusive. An ANN consists of a collection of processing elements that are highly interconnected. Given a set of inputs and a set of desired outputs, the transformation from input to output is determined by the weights associated with the inter-connections among processing elements. By modifying these interconnections, the network adapts to desired outputs. The ability of high tolerance for learning-by-example makes neural networks flexible and powerful in IDS. This paper has implemented Echo state neural network and Radial basis function applied to intrusion detection. The scope of the work includes using the available KDD database.

Keywords

Radial Basis Function (RBF) Networks Echo State Neural Networks (ESNN), KDD Features, Intrusion Detection.

Full Text

     

P. Pramila ¹, S. Purushothaman ²

Affiliations
1 Dept. of Electrical & Electronics Engg., Bangalore Institute of Technology, V.V Puram, K.R. Road, Bangalore-560004, IN
2 Department of Mechanical Engineering, PET Engineering College, Vallioor 627117, IN

Abstract

The paper presents implementation of neurofuzzy methods for estimating power quality. The electrical signal is collected and decomposed to obtain features. The estimated features were used as input parameters to Radial Basis Function (RBF) and Fuzzy Logic (FL) for training and testing to get the final weights. The resulted final weights were used for testing the proposed algorithms to estimate the power quality of the electrical signal.

Keywords

Radial Basis Function (RBF), Fuzzy Logic (FL), Power Quality Analysis.

Full Text

     

C. Berin Jones ¹, S. Suresh Kumar ², S. Purushothaman ³

Affiliations
1 Manonmaniam Sundaranar University 627012, IN
2 Vivekanandha College of Technology for Women, Tiruchencode-637205, IN
3 PET Engineering College, Vallioor-627117, IN

Abstract

This paper presents the implementation of back propagation algorithm (BPA), Fuzzy logic (FL) and graylevel co-occurrence matrix (GLCM) in identifying the exudates in diabetic retinopathy (DR) images. Human eyes are affected due to malnutrition and other present day exposure of eyes to different environments as continued work on the computer, watching television, watching small screen sized mobile phones. The eyes are strained in one form or other and damage to the nerves of the eyes occur which can be called DR, glaucoma and many other types. Representative features are obtained from the image. They are used for training the implemented algorithms. The performance of the three algorithms in identifying the exudates are presented.

Keywords

Gray Level Co-Occurrence Matrix (GLCM), Diabetic Retinopathy, Fundus Image, Artificial Neural Network (ANN), Fuzzy Logic (FL), Back Propagation Algorithm (BPA).

Full Text

     

P. Guhan ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 Department of MCA, VELS University, Chennai–600 117, IN
2 PET Engineering College, Vallioor, 627117, IN
3 Mother Teresa Women's University, Kodaikanal-624102, IN

Abstract

Fingerprint is a unique facility which is present in human anatomy. The ups and downs of the curvature present in the finger among human are different. The curvature present among male and female are also different. In general, the image of a finger either a thumb or index finger is scanned by a compact fingerprint scanner with high resolution. The fingerprint scanned w412ill go through preprocessing followed by wavelet decomposition. This paper implements wavelet decomposition for extracting features of fingerprint images. Subsequently, at the 5th level decomposition, statistical features are computed from the coefficients of approximation and detail. These features are used to train the radial basis function (RBF) neural network for identifying fingerprints. Sample finger prints are taken from database from the internet resource. The fingerprints are decomposed using daubauchi wavelet 1(db1) into 5 levels. The coefficients of approximation at the 5thlevel are used for calculating statistical features. These statistical features are used for training the RBF network.

Keywords

Fingerprint, Daubauchi Wavelet, Subband Wavelet Coefficients, Approximation and Details of 5 Level Decomposition, Radial Basis Function (Rbf).

Full Text

     

P. Sripriya ¹, S. Purushothaman ², R. Rajeswari ³

Affiliations
1 Department of MCA, VELS University, Pallavaram, Chennai-600117, IN
2 PET Engineering College, Vallioor, 627117, IN
3 Mother Teresa Women's University, Kodaikanal-624101, IN

Abstract

This paper presents the combined implementation of radial basis function(RBF) along with hidden Markov model (HMM) for human activity recognition. Surveillance cameras are installed in the crowded area in major metropolitan cities in various countries. Sophisticated algorithms are required to identify human walking style to monitor any unwanted behavior that would lead to suspicion. This paper presents the importance of RBF to identify the human GAIT. GAIT is one of the biometrics that can be measured at a distance and useful for security surveillance and biometric applications. The attraction of using GAIT as a biometric is that it is non-intrusive and typifies the motion characteristics specific to an individual. The proposed system attempt to recognize people by modeling each individual's GAIT using HMM. The HMM is a good choice for modeling a walk cycle because it can model sequential processes. This knowledge is used to generate a lower dimensional observation vector sequence which is then used to design a continuous density HMM for each individual

Keywords

GAIT, Human Walking Action, Radial Basis Function, Hidden Markov Model.

Full Text

     

R. I. Rajidap Neshtar ¹, S. Purushothaman ²

Affiliations
1 Department of Mechanical Engineering, Vinayaka Missions University, Salem, IN
2 Department of Mechanical Engineering, PET Engineering College, Tirunelveli District-627117, IN

Abstract

This paper presents estimation of stress and strain of a Rapid prototype product using artificial neural network (ANN). Radial basis function network is used to train the ANN topology. 3D model of blade is developed by using PROE. The model is analyzed using ANSYS to find the Von Mises stress and equivalent strain. The algorithm is trained using 15 values in the input layer of the ANN topology and two values in the output layer: stress and strain that are to be estimated during the testing stage of RBF algorithm.

Keywords

Radial Basis Function Network, Finite Element Method, Structural Analysis, and Blade.

Full Text

     

M. Suresh ¹, Maheswar Dutta ², S. Purushothaman ³

Affiliations
1 Mechanical Engineering, Sri Sai Ram Engineering College, Chennai-44, IN
2 M.N.R Engineering College, Hyderabad, IN
3 Mechanical Engineering, Udaya School of Engineering, IN

Abstract

This paper presents the implementation of functional back propagation algorithm (FUBPA) for estimating the power, torque, specific fuel consumption and presence of carbon monoxide, hydrocarbons in the emission of a direct injection diesel engine. Experimental readings were obtained using the biodiesel prepared from the waste cooking oil collected from the canteen of Sri Sairam Engineering College, India. This waste cooking oil was due to the preparation of varieties of food (vegetables fried and non vegetarian). To obtain the biodiesel, transesterification was done in chemical lab for more than a week, and the biodiesel was obtained. The biodiesel was mixed in proportions of 10%, 20%, 30%, 40%, 50% with remaining combinations of the diesel supplied by the Indian government. Variable compression ratio (VCR) diesel engine with single cylinder, 4 stroke diesel type was used.
The outputs of the engine as power, torque and specific fuel consumption were obtained from the computational facility attached to the engine. The data collected for different input conditions of the engine was further used to train FUBPA.
The trained FUBPA network was further used to predict the power, torque and SFC for different speed, biodiesel and diesel combinations and full load conditions. The estimation performance of the FUBPA network is discussed.

Keywords

Functional Back Propagation Algorithm, Waste Cooking Oil, Biodiesel.

Full Text

     

N. Austin ¹, P. Senthilkumar ², S. Purushothaman ³

Affiliations
1 Satyabama University, IN
2 Department of Mechanical Engineering, KSR College of Engineering, Tiruchengode, IN
3 Udaya School of Engineering, Kanyakumari District-629204, IN

Abstract

This paper presents implementation of Radial basis function (RBF) neural network to find out mixture of Hydrofluorocarbon (HFC) and Hydrocarbon (HC) for obtaining higher Coefficients of Performances (COPs). The thermodynamic properties of refrigerants are obtained using REFPROP 9 software that contains details of refrigerants. Different combinations of the refrigerants along with their COPs are obtained by the REFPROP 9. It consumes time in obtaining the correct combination of refrigerants as lot of menu options have to be chosen in the REFPROP 9. In order to make the process of finding out the correct mixed refrigerants with less manual intervention, RBF is trained and tested with the patterns of mixed refrigerants. The RBF mixed refrigerant analysis software has been developed by using MATLAB 10.

Keywords

Radial Basis Function, Artificial Neural Network, Mixed Refrigerant, Coefficient of Performance.

Full Text

     

P. Radha ¹, E. Ramaraj ², S. Purushothaman ³

Affiliations
1 Department of Computer Science and Engineering, Alagappa University, Karaikudi, IN
2 Computer Center, Alagappa University, Karaikudi, IN
3 Udaya School of Engineering, 629204, IN

Abstract

This paper discusses the preprocessing and classification of depression data using back propagation algorithm (BPA). In general, input vectors will not be orthogonal to each other. The proposed method of preprocessing the input vector makes possible BPA learn the input vectors. The classification performance of BPA have been shown for a minimum 80%.

Keywords

Depression Data, Back Propagation Algorithm, Polynomial Discriminant Vector (PDV).

Full Text

     

R. Bhuvana ¹, S. Purushothaman ², P. Rajeswari ³

Affiliations
1 Department of Computer Science, A. M. Jain College, Chennai - 600 114, Tamil Nadu, IN
2 Institute of Technology, Haramaya University, ET
3 Department of ECE, Institute of Technology, Haramaya University, ET

Abstract

Depression is common in working personality due to high tension in the working environment. Symptoms can affect dayto- day life and can become very worrying. With true depression, there is a low mood and other symptoms each day for more than two weeks. Symptoms can also become rigorous enough to interfere with normal day-to-day activities. This paper suggests an Artificial Neural Network (ANN) algorithm approach for quicker learning of psychological depression data. Performance of neural network methods for estimating depression state with Echo State Neural Network (ESNN) is presented. Tentative data were collected from the patients with 21 input variables. One target output is used for training the ESNN. The training and testing patterns are made using the data as per Hamilton Rating Scale.The input patterns are pre-processed and presented to the input layer of ANN. The proposed method proves to be a capable system for diagnosis of depression.

Keywords

Depression Data, ESNN, Hamilton Rating Scale

Full Text