K. V. Krishna Kishore

Prediction of student academic progression: A case study on Vignan University

To explore the academic progression of the students, higher educational institutions need better assessment and prediction tools. In this regard, Multilayer Perceptron (MLP) based prediction application is proposed to predict the Grade Point Average (GPA) of the Undergraduate students by the make use of students Previous Academic History, Regularity, No. of Backlogs, Degree of Intelligence, Working Nature, Discipline, Social Activities and Grade. With this application it is possible to predict the students data that who are at risk, and some proactive measures like extra classes & supporting material are offered to improve the academic progress of those students. To evaluate the performance of the proposed application, data has recorded from 134 third Year Computer Science Engineering Students of Vignan University and achieved 95.52% and 97.37% of prediction accuracy with RBF and MLP respectively.

Performance estimation of DSR, DSDV and AODV in TCP, UDP and SCTP

Mobile ad-hoc network is a group of mobile nodes without infrastructure and self-configured in nature. Each node in MANET can move freely in Omni direction thus causes change in links to other devices frequently and they must forward packet transmission irrespective of its personal use; because of this a router is required. So the three prominent routing protocols in MANET such as DSR, AODV and DSDV are considered for the routing in MANETs. In this paper the comparison is based on the performance of two reactive routing protocols DSR, AODV and the proactive DSDV protocol in TCP, UDP and SCTP environments. A simulation model MAC and 802.11 is used to study the interlayer interactions. The performance of protocols is measured based on packet loss and delay. The results are carried out by using network simulator-2. The results presented in this paper illustrate the importance and evaluation of the routing protocols.

An efficient multimodal person authentication system using Gabor and sub band coding

This paper presents fusion of two biometric traits, i.e., face and speech, at matching score level fusion. The features are extracted from the pre-processed images of face and speech. Gabor Wavelet and Discrete Cosine Transform (DCT) are used to extract facial features and Sub Band Coding (SBC) is used to extract features from speech signals. These features of a probe image are compared with training images of each trait and then calculate matching score. The individual scores generated after each matching are passed to the fusion module. The final score is then used to declare the person as genuine or an impostor. The proposed method is tested on ORL database and it outperforms with False Acceptance Rate of 0.75% and False Rejection Rate of 1.24%.

Facial expression classification using Kernel based PCA with fused DCT and GWT features

Although many approaches for facial expression recognition have been proposed in the past, most of them yielding poor recognition performance with single feature extraction method. The objective of this paper is to propose an innovative method based on fusion of local and global features for better classification rate. Gabor wavelets(GWT) are used to extract Local features and Discrete Cosine Transform (DCT) is used to extract global features from facial expression images. To reduce dimensionality of extracted features and better classification performance Kernel Principal Components Analysis (KPCA) is applied. Wavelet fusion method is used to fuse the features extracted from GWT and DCT. Finally the images are classified into 6 different basic emotions like surprise, fear, sad, joy, anger and disgust using Radial Basis Function(RBF) Neural Network classifier. The performance of the proposed method is evaluated on Cohn-Kanade database. The results of proposed algorithm exhibit high performance rate of about 99% in person dependent facial expression recognition.

Hybrid Face Recognition with Locally Discriminating Projection

The face recognition task involves extraction of unique features from the human face. Manifold learning methods are proposed to project the original data into a lower dimensional feature space by preserving the local neighborhood structure. LPP should be seen as an alternative to Principal Component Analysis (PCA). When the high dimensional data lies on a low dimensional manifold embedded in the ambient space, the Locality Preserving Projections are obtained by doing the optimal linear approximations to the Eigen functions of the Laplace Beltrami operator on the manifold. However, LPP is an unsupervised feature extraction method because it considers only class information. LDP is the recently proposed feature extraction method different from PCA and LDA, which aims to preserve the global Euclidean structure, LDP is the extension of LPP, which seeks to preserve the intrinsic geometry structure by learning a locality preserving submanifold. LDP is a supervised feature extraction method because it considers both class and label information. LDP performs much better than the other feature extraction methods such as PCA and Laplacian faces. In this paper LDP along with Wavelet features is proposed to enhance the class structure of the data with local and directional information. In this paper, the face Image is decomposed into different subbands using the discrete wavelet transform bior3.7, and the subbands which contain the discriminatory information are used for the feature extraction with LDP. In general the size of the face database is too high and it needs more memory and needs more time for training so that to improve time and space complexities there is a need for dimensionality reduction. It is achieved by using both biorthogonal wavelet transform and LDP the features extracted take less space and take low time for training. Experimental results on the ORL face Database suggests that LDP with DWT provides better representation and achieves lower error rates than LDP with out wavelets and has lower time complexity. The subband faces performs much better than the original image in the presence of variations in lighting, and expression and pose. This is because the subbands which contain discriminatory information for face recognition are selected for face representation and others are discarded.

SVM kernel based predictive analytics on faculty performance evaluation

In recent years, higher education has been gaining importance in graduate students to make successful careers. So, academic organizations are given utmost importance for quality in academics to build the careers of the students. Faculty performance plays a vital role in academic institutions. In this paper, the performance of faculty members is evaluated on the basis of different parameters are taken for assessment and predicted by building models using data mining techniques. In this evaluation, the sample data is collected, preprocessed, and model learning is done using Support vector machines (SVM) with several kernel methods such as linear, sigmoid, radial basis, polynomial and Pearson VII function-based universal kernel (PUKF). The idea of this proposed paper is to investigate and analyze by considering various parameters for predicting the performance of faculty. The parameters considered are Faculty profile, Quality of Teaching, Maintaining Relationships, Learning Assessment, Counseling and Mentoring, Administrative Functions, Research and Development, Organizational Qualities and Outcome. Performance of various kernels is evaluated with the data and models with SVM-PUKF yields better accuracy by 97.84% when compared with other three standard kernels.

Person identification system using feature level fusion of multi-biometrics

The use of unimodal biometric system is very low because of physiological defects, modes of user and their environment. Some of those drawbacks are alleviated by providing same identity for multiple evidences. Here a multimodal biometric system is proposed based on LBP, PCA and probabilistic neural network (PNN). In proposed method LBP extracted the Face features from face images and those features are given as input to PCA that generates Face Feature Vector with reduced Dimensions. Finger features are extracted from Fingerprint images using LBP and those features are given as input to PCA that generates Finger Feature Vector with reduced Dimensions. Using LBP, the distinct textual features of face and fingerprint are extracted. Weighted Summation Fusion method is used to combine these unimodal features. A probabilistic neural network is used as Classifier. An average recognition rate of 97.5% achieved with proposed method. Proposed method show that the proposed algorithm requires low computational cost.

An efficient text dependent speaker recognition using fusion of MFCC and SBC

In this paper an efficient approach for the recognition of a speaker based on text dependent speech is presented. Speaker Recognition/ Verification system suffers with wide variety of problems. In the proposed approach, the features are extracted using two methods such as Mel Frequency Cepstral Coefficients and wavelet subband coefficients, and then these futures are fused through concatenation to give optimum performance. Those concatenated feature set are more reliable to discriminate an imposter from the genuine. Those concatenated features are classified using support vector machine classifier. Performance of the proposed approach is validated on a self generated corpus of size 300 samples of 20 individual. The proposed method outperforms other existing methods.

A novel E-learning framework to learn IT skills for visual impaired

In this paper a novel approach is implemented for learning of IT related courses for visually impaired through e-learning which provides a facility for anyone to learn from anywhere at anytime. It enables visually impaired students to learn the knowledge by accessing the e-content and work more independently with ease of access compared to traditional methods. At present most of eLearning websites are providing the content mostly useful for normal students. Using screen readers, learning of the content from these websites are cumbersome for visual impaired. A survey has been carried out on a batch of visually impaired students to explore existing eLearning concepts and tools in IT related courses. This work is mostly focused on the issue of web accessibility and to facilitate effective access of content and learning to the visually-impaired students. The methodology used in this paper is validated by visually-impaired students and experts and yielded good results. It is also tested for portability of using with various modes of access such as personal computers, Tablets and Mobile phones.

Hybrid framework for face recognition with expression & illumination variations

In this paper, a hybrid framework is proposed to improve the performance of face recognition by combining global descriptors and local appearance descriptors and proved that their complementary nature makes them good candidates in the better recognition of faces. The proposed face recognition method can handle facial appearance variations which are caused by facial expression and illumination under controlled capture conditions. Different from traditional face recognition methods, the proposed method uses multiple features which are extracted using Global and Local feature extraction algorithms like Principal Component Analysis (PCA) & Local Binary Pattern (LBP). Wavelet fused feature vector has richer information than feature vector extracted using unifeature extraction algorithms. Radial Basis Function (RBF) is used to classify feature vectors. The proposed method has been extensively evaluated on the standard benchmark databases like ORL and Grimace. It is found that significant results obtained in comparison with well-known generic face recognition methods.