Browsing by Author "Radha, N"

Now showing 1 - 18 of 18

BAYESIAN CLASSIFICATION FOR IMAGE RETRIEVAL USING VISUAL DICTIONARY
(Springer International Publishing, 2013) Nazirabegum, M K; Radha, N
Image Retrieval is one of the most promising technologies for retrieving images through the query image. It enables the user to search for the images based upon the relevance of the query image. The main objective of this paper is to develop a faster and more accurate image retrieval system for a dynamic environment such as World Wide Web (WWW). The image retrieval is done by considering color, texture, and edge features. The bag-of-words model can be applied to image classification, by treating image features as words. The goal is to improve the retrieval speed and accuracy of the image retrieval systems which can be achieved through extracting visual features. The global color space model and dense SIFT feature extraction technique have been used to generate a visual dictionary using Bayesian algorithm. The images are transformed into set of features. These features are used as an input in Bayesian algorithm for generating the code word to form a visual dictionary. These code words are used to represent images semantically to form visual labels using Bag-of-Features (BoF). Then it can be extended by combining more features and their combinations. The color and bitmap method involves extracting only the local and global features such as mean and standard deviation. But in this classification technique, color, texture, and edge features are extracted and then Bayesian Algorithm is applied on these image features which gives acceptable classification in order to increases the accuracy of image retrieval.
BAYESIAN CLASSIFICATION FOR IMAGE RETRIEVAL USING VISUAL DICTIONARY
(Springer Link, 2013) Nazirabegum, M K; Radha, N
Image Retrieval is one of the most promising technologies for retrieving images through the query image. It enables the user to search for the images based upon the relevance of the query image. The main objective of this paper is to develop a faster and more accurate image retrieval system for a dynamic environment such as World Wide Web (WWW). The image retrieval is done by considering color, texture, and edge features. The bag-of-words model can be applied to image classification, by treating image features as words. The goal is to improve the retrieval speed and accuracy of the image retrieval systems which can be achieved through extracting visual features. The global color space model and dense SIFT feature extraction technique have been used to generate a visual dictionary using Bayesian algorithm. The images are transformed into set of features. These features are used as an input in Bayesian algorithm for generating the code word to form a visual dictionary. These code words are used to represent images semantically to form visual labels using Bag-of-Features (BoF). Then it can be extended by combining more features and their combinations. The color and bitmap method involves extracting only the local and global features such as mean and standard deviation. But in this classification technique, color, texture, and edge features are extracted and then Bayesian Algorithm is applied on these image features which gives acceptable classification in order to increases the accuracy of image retrieval.
BODY JOINTS AND TRAJECTORY GUIDED 3D DEEP CONVOLUTIONAL DESCRIPTORS FOR HUMAN ACTIVITY IDENTIFICATION
(Blue Eyes Intelligence Engineering & Sciences Publication, 2019-10) Srilakshmi, N; Radha, N
Human Activity Identification (HAI) in videos is one of the trendiest research fields in the computer visualization. Among various HAI techniques, Joints-pooled 3D-Deep convolutional Descriptors (JDD) have achieved effective performance by learning the body joint and capturing the spatiotemporal characteristics concurrently. However, the time consumption for estimating the locale of body joints by using large-scale dataset and computational cost of skeleton estimation algorithm were high. The recognition accuracy using traditional approaches need to be improved by considering both body joints and trajectory points together. Therefore, the key goal of this work is to improve the recognition accuracy using an optical flow integrated with a two-stream bilinear model, namely Joints and Trajectory-pooled 3D-Deep convolutional Descriptors (JTDD). In this model, an optical flow/trajectory point between video frames is also extracted at the body joint positions as input to the proposed JTDD. For this reason, two-streams of Convolutional 3D network (C3D) multiplied with the bilinear product is used for extracting the features, generating the joint descriptors for video sequences and capturing the spatiotemporal features. Then, the whole network is trained end-to-end based on the two-stream bilinear C3D model to obtain the video descriptors. Further, these video descriptors are classified by linear Support Vector Machine (SVM) to recognize human activities. Based on both body joints and trajectory points, action recognition is achieved efficiently. Finally, the recognition accuracy of the JTDD model and JDD model are compared.
CANCELLABLE MULTIMODAL BIOMETRIC USER AUTHENTICATION SYSTEM WITH FUZZY VAULT
(IEEE, 2016-05-30) Soruba Sree, S R; Radha, N
Biometrics refers to authentication techniques that rely on humans physical and behavioral characteristics that can be automatically checked. Biometric based authentication system provides robust security and ease of use than conventional methods of verification system. Multimodal biometric system is one of the major areas of study identified with large applications in recognition system. Unimodal biometric systems challenge with a wide variety of problems such as noisy data, Intra-class variations, non-universality, and spoof attacks. Some of these limitations can be solved in multimodal biometric system. In proposed work, face and fingerprint biometric traits are used for multimodal biometric authentication system. Biometric traits are transformed using distortion algorithm. After the transformation processes pre-processing of images are done to improve the clear visibility of images. The extractions of minutiae features from fingerprint are achieved using Crossing Number concept and the face features are extracted using the Local Binary Pattern algorithm. To combine both the face and fingerprint features feature level fusion is used. In order to provide additional security to the proposed work the fuzzy vault is introduced by adding duplicate values and having a secret key to lock and unlock the system. Fuzzy vault and distortion acts as an additional layer of security in multimodal biometric user authentication system.
DEEP POSITIONAL ATTENTION-BASED BIDIRECTIONAL RNN WITH 3D CONVOLUTIONAL VIDEO DESCRIPTORS FOR HUMAN ACTION RECOGNITION
(IOP Publishing Ltd, 2021) Srilakshmi, N; Radha, N
This article presents the Joints and Trajectory-pooled 3D-Deep Positional Attention-based Bidirectional Recurrent convolutional Descriptors (JTPADBRD) for recognizing the human activities from video sequences. At first, the video is partitioned into clips and these clips are given as input of a two-stream Convolutional 3D (C3D) network in which the attention stream is used for extracting the body joints locations and the feature stream is used for extracting the trajectory points including spatiotemporal features. Then, the extracted features of each clip is needed to aggregate for creating the video descriptor. Therefore, the pooled feature vectors in all the clips within the video sequence are aggregated to a video descriptor. This aggregation is performed by using the PABRNN that concatenates all the pooled feature vectors related to the body joints and trajectory points in a single frame. Thus, the convolutional feature vector representations of all the clips belonging to one video sequence are aggregated to be a descriptor of the video using Recurrent Neural Network (RNN)-based pooling. Besides, these two streams are multiplied with the bilinear product and end-to-end trainable via class labels. Further, the activations of fully connected layers and their spatiotemporal variances are aggregated to create the final video descriptor. Then, these video descriptors are given to the Support Vector Machine (SVM) for recognizing the human behaviors in videos. At last, the experimental outcomes exhibit the considerable improvement in Recognition Accuracy (RA) of the JTDPABRD is approximately 99.4% achieved on the Penn Action dataset as compared to the existing methods.
MULTI-VIEW HUMAN ACTION RECOGNITION USING ADAPTIVE OPTIMIZATION ALGORITHM WITH SKELETON BASED GRAPH NEURAL NETWORKS (Conference Paper)
(Institute of Electrical and Electronics Engineers Inc., 2024-01-25) Srilakshmi, N; Radha, N
Recognizing human actions from multi-viewpoint video data is a complex task with applications in various fields, including surveillance, robotics, and human-computer interaction. An innovative approach is presented in this study to Multi-View Human Action Recognition (MV-HAR) using an Adaptive Optimization Algorithm (AOA) combined with a Skeleton-based Graph Neural Network (SGNN). The proposed architecture aims to leverage the complementary information from multiple viewpoints while effectively capturing the temporal and spatial dependencies inherent in human actions. The pre-processing pipeline aligns and fuses skeleton data extracted from different viewpoints, producing a coherent representation of the action across cameras. Each skeleton sequence is transformed into a graph structure, where joints represent nodes and edges encapsulate the relationships between joints over time. This sequence of graphs is then processed by the SGNN, which learns to capture the evolving dynamics of the action through multiple graph convolutional layers. On benchmark datasets, the proposed approach proves effective with an accuracy of 96.8%.
MULTIMODAL BIOMETRIC TEMPLATE AUTHENTICATION OF FINGER VEIN AND SIGNATURE USING VISUAL CRYPTOGRAPHY
(IEEE, 2016-05-30) Nandhinipreetha, A; Radha, N
In this paper personal verification method using finger-vein and signature is presented. Among many authentication systems finger-vein is promising as the foolproof method of automatic personal identification. Finger-vein and signature image is pre-processed and features are extracted using cross number concept and principle compound analysis. Fusion technique is used to fuse the finger vein and signature images. Then the visual cryptographic scheme is applied for the biometric template to generate the shares. The shares are stored in a separate database, and then the biometric image is revealed only when both the shares are simultaneously available. At the same time, the individual image does not reveal the identity of the biometric image. The proposed work is evaluated with evaluation metrics FAR, FRR and accuracy.
A NEW FRAMEWORK FOR IRIS AND FINGERPRINT RECOGNITION USING SVM CLASSIFICATION AND EXTREME LEARNING MACHINE BASED ON SCORE LEVEL FUSION
(IEEE, 2013-03-21) Sangeetha, S; Radha, N
In a Multimodal biometric system, the effective fusion method is necessary for combining information from various single modality systems. Two biometric characteristics are considered in this study: iris and fingerprint. Multimodal biometric system needs an effective fusion scheme to combine biometric characteristics derived from one or more modalities. The score level fusion is used to combine the characteristics from different biometric modalities. Fusion at the score level is a new technique, which has a high potential for efficient consolidation of multiple unimodal biometric matcher outputs. Support vector machine and extreme learning techniques are used in this system for recognition of biometric traits. In this, the Fingerprint-Iris system provides better performance and comparison of support vector machine and extreme learning machine based on score-level fusion methods is obtained. In score-level fusion, ELM provides better performance as compare to the SVM. It reduces the classification time of current system. This work is valuable and makes an efficient accuracy in such applications. This system can be utilized for person identification in several applications.
PERFORMANCE ANALYSIS OF ABSTRACT-BASED CLASSIFICATION OF MEDICAL JOURNALS USING MACHINE LEARNING TECHNIQUES
(Springer Link, 2021-09-14) Deepika, A; Radha, N
Researchers face many challenges in finding the opt web-based resources by giving the queries based on keyword search. Due to advent of Internet, there are huge biological literatures that are deposited in the medical database repository in recent years. Nowadays, as many web-based medical researchers evolved in the field of medicine, there is need for an intelligent and efficient extraction technique required to filter appropriate and opt literature from the growing body of biomedical literature repository. In this research work, new combination of model is proposed in order to find the new insights in applying the combination of algorithm on biological data set. The information in the biomedical field is the basic information for healthy living. National Center for Biotechnology Information (NCBI)’s PubMed is the major source of peer-reviewed biomedical documents for researchers and health practitioners in the field of health-related management. In this paper, abstracts available in PubMed database is used for experimentation. In recent years, deep learning-based neural approach models provide an efficient way to create an end-to-end model that can accurately measure classification labels. This research work is a systematic analysis of performance of the supervised learning models such as Naïve Bayes (NB), support vector machine (SVM) and long short-term memory (LSTM) by implementing on textual medical data. The novelty in this work is the process of incorporating certain topic modelling techniques after the pre-processing phase to automatically label the documents. Topic modelling is a useful technique in increasing the efficiency and improves the ability of researchers to interpret biological information. So, the classification algorithms thus proposed are implemented in combination with popular topic modelling algorithms such as latent Dirichlet algorithm (LDA) and non-negative matrix factorization (NMF). The final performance of the combination of algorithms is also analysed and is found that SVM with NMF outperforms the other models.
PERFORMANCE ANALYSIS OF ABSTRACT-BASED CLASSIFICATION OF MEDICAL JOURNALS USING MACHINE LEARNING TECHNIQUES
(Springer Link, 2021-09-14) Deepika, A; Radha, N
Researchers face many challenges in finding the opt web-based resources by giving the queries based on keyword search. Due to advent of Internet, there are huge biological literatures that are deposited in the medical database repository in recent years. Nowadays, as many web-based medical researchers evolved in the field of medicine, there is need for an intelligent and efficient extraction technique required to filter appropriate and opt literature from the growing body of biomedical literature repository. In this research work, new combination of model is proposed in order to find the new insights in applying the combination of algorithm on biological data set. The information in the biomedical field is the basic information for healthy living. National Center for Biotechnology Information (NCBI)’s PubMed is the major source of peer-reviewed biomedical documents for researchers and health practitioners in the field of health-related management. In this paper, abstracts available in PubMed database is used for experimentation. In recent years, deep learning-based neural approach models provide an efficient way to create an end-to-end model that can accurately measure classification labels. This research work is a systematic analysis of performance of the supervised learning models such as Naïve Bayes (NB), support vector machine (SVM) and long short-term memory (LSTM) by implementing on textual medical data. The novelty in this work is the process of incorporating certain topic modelling techniques after the pre-processing phase to automatically label the documents. Topic modelling is a useful technique in increasing the efficiency and improves the ability of researchers to interpret biological information. So, the classification algorithms thus proposed are implemented in combination with popular topic modelling algorithms such as latent Dirichlet algorithm (LDA) and non-negative matrix factorization (NMF). The final performance of the combination of algorithms is also analysed and is found that SVM with NMF outperforms the other models.
PERFORMANCE ANALYSIS OF DEEP LEARNING ALGORITHMS FOR HUMAN ACTION RECOGNITION USING SPATIO TEMPORAL FEATURES FROM VIDEO IMAGES
(2023) Sri Lakshmi, N; Radha, N
newline
SECURE AND ATTACK AWARE ROUTING IN MOBILE AD HOC NETWORKS AGAINST WORMHOLE AND SINKHOLE ATTACKS
(IEEE, 2017-10) Sasirekha, D; Radha, N
The inherent characteristics of Mobile Ad hoc network (MANET) such as dynamic topology, limited bandwidth, limited power supply, infrastructure less network make themselves attractive for a wide spectrum of applications and vulnerable to security attacks. Sinkhole attack is the most disruptive routing layer attack. Sinkhole nodes attract all the traffic towards them to setup further active attacks such as Black hole, Gray hole and wormhole attacks. Sinkhole nodes need to be isolated from the MANET as early as possible. In this paper, an effective mechanism is proposed to prevent and detect sinkhole and wormhole attacks in MANET. The proposed work detects and punishes the attacker nodes using different techniques such as node collusion technique, which classifies a node as an attacker node only with the agreement with the neighboring nodes. When the node suspects the existence of attacker or sinkhole node in the path, it joins together with neighboring nodes to determine the sinkhole node. In the prevention of routing attacks, the proposed system introduces a route reserve method; new routes learnt are updated in the routing table of the node only after ensuring that the route does not contain the attacker nodes. The proposed system effectively modifies Ad hoc on demand Distance Vector (AODV) with the ability to detect and prevent the sinkhole and wormhole attack, so the modified protocol is named as Attack Aware Alert (A3AODV). The experiments are carried out in NS2 simulator, and the result shows the efficiency in terms of packet delivery ratio and routing overhead.
SECURING DATA TRANSMISSION IN MANET USING AN IMPROVED COOPERATIVE BAIT DETECTION APPROACH
(IEEE Xplore, 2017-03-30) Nachammai, M; Radha, N
Mobile Ad hoc Network often called as MANET, which does not have any particular infrastructure, in which each mobile devices are connected wirelessly and can move freely in any direction without having any restrictions in the network. Malicious nodes present in this network can easily launch highly vulnerable attacks like collaborative Black hole attack and Gray hole attack due to its dynamically changing topology. These attacks affect the routing process within MANET. Hence, Security is the primary concern for finding these nodes. But, to prevent or detect malicious nodes that causes Gray hole or a collaborative black hole attack is a challenge. In this scheme, the malicious nodes and its behaviours are detected using reverse tracing technique by sending RREQ and RREP. However security for transmitting data is not considered by CBDS. In order to have secure transmission after the malicious node detection, our proposed system uses an improved Cooperative Bait Detection approach which incorporates CBDS with message security schemes. Finally this approach is compared with the existing system by using performance metrics like End-to-End Delay, Packet Delivery Ratio (PDR), Throughput and Routing Overhead.
SECURING IRIS AND FINGERPRINT TEMPLATES USING FUZZY VAULT AND SYMMETRIC ALGORITHM
(IEEE, 2013-03-21) Sowkarthika, S; Radha, N
The important aspect of all verification system is authentication and security. This aspect necessitates the development of a method that ensures user security and privacy. The traditional methods such as tokens and passwords provide security to the users. Uncertainly, the attackers can easily compromise these techniques. In recent years, the combination of biometrics and cryptography techniques has been proved as a efficient way to achieve security. The important feature of using biometric template is that it cannot be exploit by an unauthorized user. Most commonly used biometric features are iris, retina, fingerprint, face, palmprint, hand geometry, voice and so on. Fuzzy vault is the concept which uses the combination of biometrics and cryptographic key generation technique. This fuzzy vault act as a additional layer of security. This paper proposes a biometric verification system investigating the combined usage of multimodal biometric features and fuzzy vault scheme. This approach uses of fingerprint and iris in order to provide higher accuracy rate. Experiments were conducted to investigate the performance of the proposed system in ensuring the user security and privacy.
SECURING PATIENT'S CONFIDIENTIAL INFORMATION USING ECG STEGANOGRAPHY
(IEEE, 2017-10) Sivaranjani, B; Radha, N
The main goal of this research work is to enhance the security of Patient's medical data. During the transmission, the data is concealed with ECG signal. The ECG signal of the human being vary from one person to person. Like other Biometric traits, ECG cannot be imitated and duplicated. Encryption is one of the best technique that guarantees the security of sensitive information. This technique not only grants the information's security but also authentication. secret sub keeping system security. digital signature, and etc. Therefore, the purpose of adopting encryption techniques is to ensure the information's integrity and confidentiality, that prevents information from tampering, forgery and counterfeiting. The encryption method used in this work will encrypt the secret data into unreadable form by creating the information inaccessible to any hacker having a random method. In this paper, Haar Wavelet Transformation is used to decompose an ECG signal to different frequency sub-bands. RSA algorithm is used to encrypt the patient information with the help of key pairs. Arnold cat map technique is used to scramble the encrypted data for more security of information. And for embedding, Singular Value Decomposition (SVD) is used for effective transformation with high security. The proposed algorithm is evaluated based on MSE, PSNR, AD, NCC, MD, NAE, BER and accuracy. The experimental results prove that the performance obtained using proposed techniques give better results than the existing techniques.
SPAM CLASSIFICATION USING SUPERVISED LEARNING TECHNIQUES
(ACM Digital Library, 2010-09-16) Deepa Lakshmi, R; Radha, N
Spam message is one of the major problems in today's Internet, which brings financial damage to companies and annoying individual users. Spam filtering is able to control the problem in a variety of ways. Many researches in spam filtering have been centered on the classifier-related issues. Currently, machine learning for spam classification is an important research issue at present. This paper explores and identifies the use of different machine learning algorithms for classifying spam messages from e-mail. Finally, a comparative analysis among the algorithms has also been presented with spam classification.
A SURVEY ON NETWORK INTRUSION SYSTEM ATTACKS CLASSIFICATION USING MACHINE LEARNING TECHNIQUES
(IOP Publishing Ltd, 2021) Deepa, V; Radha, N
Wireless Local Area Network (WLAN) security management is now being confronted by rapid expansion in wireless network errors, flaws and assaults. In recent times, as computers are used extensively through network and application creation on numerous platforms, attention is provided to network security. This definition includes security vulnerabilities in both complicated and costly operating programs. Intrusion is also seen as a method of breaching security, completeness and availability. Intrusion Detection System (IDS) is an essential method for the identification of network security vulnerabilities and abnormalities. A variety of significant work has been carried out on intrusion detection technologies often seen as premature not as a complete method for countering intrusion. It has also become a most challenging and priority tasks for security experts and network administrators. Hence, it cannot be replaced by more secure systems. Data mining used for IDS can effectively identify intrusion and the identified intrusion values are used to predict further intrusion in future. This paper presents a detailed review of literature about how data mining techniques were utilized for intrusion detection. First, intrusion detection on various benchmark and real-time datasets by data mining techniques are studied in detail. Then, comparative study is conducted with their merits and demerits for identifying the challenges in those techniques and then this paper is concluded with suggestions of solutions for enhancing the efficiency of intrusion detection in the network.
A SURVEY ON NETWORK INTRUSION SYSTEM ATTACKS CLASSIFICATION USING MACHINE LEARNING TECHNIQUES
(IOP Publishing Ltd, 2021) Deepa, V; Radha, N
Wireless Local Area Network (WLAN) security management is now being confronted by rapid expansion in wireless network errors, flaws and assaults. In recent times, as computers are used extensively through network and application creation on numerous platforms, attention is provided to network security. This definition includes security vulnerabilities in both complicated and costly operating programs. Intrusion is also seen as a method of breaching security, completeness and availability. Intrusion Detection System (IDS) is an essential method for the identification of network security vulnerabilities and abnormalities. A variety of significant work has been carried out on intrusion detection technologies often seen as premature not as a complete method for countering intrusion. It has also become a most challenging and priority tasks for security experts and network administrators. Hence, it cannot be replaced by more secure systems. Data mining used for IDS can effectively identify intrusion and the identified intrusion values are used to predict further intrusion in future. This paper presents a detailed review of literature about how data mining techniques were utilized for intrusion detection. First, intrusion detection on various benchmark and real-time datasets by data mining techniques are studied in detail. Then, comparative study is conducted with their merits and demerits for identifying the challenges in those techniques and then this paper is concluded with suggestions of solutions for enhancing the efficiency of intrusion detection in the network.