Semi-Supervised Learning on Riemannian Manifolds

We consider the general problem of utilizing both labeled and unlabeled data to improve classification accuracy. Under the assumption that the data lie on a submanifold in a high dimensional space, we develop an algorithmic framework to classify a partially labeled data set in a principled manner. The central idea of our approach is that classification functions are naturally defined only on the submanifold in question rather than the total ambient space. Using the Laplace-Beltrami operator one produces a basis (the Laplacian Eigenmaps) for a Hilbert space of square integrable functions on the submanifold. To recover such a basis, only unlabeled examples are required. Once such a basis is obtained, training can be performed using the labeled data set. Our algorithm models the manifold using the adjacency graph for the data and approximates the Laplace-Beltrami operator by the graph Laplacian. We provide details of the algorithm, its theoretical justification, and several practical applications for image, speech, and text classification.     

Formulation and in Vitro — In Vivo Characterization of Solid Dispersions of Piroxicam

The present work embodies studies performed with solid dispersions of the non-steroidal anti-inflammatory agent piroxicam, using biocompatible water soluble polymers polyethylene glycol (PEG) 4000 and 6000 alone, as well as in blends of various proportions. Different physicochemical properties and in vitro characterization were carried out. In vivo studies were performed to correlate with in vitro studies.     

On the stability of intermetallic phases

A general methodology using atomic clusters is applied to three problems connected to the study of alloy phase stability. The cluster method proposed by Allen and Cahn is applied to non-ideal hcp structures under tetrahedral approximation using multiatom interactions. The possible ground-state structures which are stable at absolute zero temperature are obtained. A geometrical representation in 4D parameter space of the possible strengths of multiatom interactions permitted for these structures is illustrated in terms of a 2D analogue. Extending these ideas, the cluster variation method (CVM) proposed by Kikuchi is applied to fcc structures under tetrahedral approximation to find the effect of multiatom interactions on the topology of the coherent phase diagrams in which all the phases present are derivable by mere rearrangement of atoms on the parent disordered structure. In addition, the possible invariant reactions are identified in such coherent phase diagrams. Finally the CVM is applied for calculating a model incoherent phase diagram, that of Ti-Zr system, where disordered hcp and bcc phases are present. The free energies of hcp and bcc phases are formulated using CVM procedures respectively under tetrahedral-octahedral and tetrahedral approximations. The CVM is shown to be in better agreement with the thermodynamic data and to be able to reproduce the correct value of measured enthalpy of transformation compared to that given by the regular solution model, which significantly overestimates the same.     

STEM: Secure Telephony Enabled Middlebox

Dynamic applications, including IP telephony, have not seen wide acceptance within enterprises because of problems caused by the existing network infrastructure. Static elements, including firewalls and network address translation devices, are not capable of allowing dynamic applications to operate properly. The Secure Telephony Enabled Middlebox (STEM) architecture is an enhancement of the existing network design to remove the issues surrounding static devices. The architecture incorporates an improved firewall that can interpret and utilize information in the application layer of packets to ensure proper functionality. In addition to allowing dynamic applications to function normally, the STEM architecture also incorporates several detection and response mechanisms for well-known network-based vulnerabilities. This article describes the key components of the architecture with respect to the SIP protocol.     

Edge structure preserving 3D image denoising by local surface approximation

In various applications, including magnetic resonance imaging (MRI) and functional MRI (fMRI), 3D images are becoming increasingly popular. To improve the reliability of subsequent image analyses, 3D image denoising is often a necessary preprocessing step, which is the focus of the current paper. In the literature, most existing image denoising procedures are for 2D images. Their direct extensions to 3D cases generally cannot handle 3D images efficiently because the structure of a typical 3D image is substantially more complicated than that of a typical 2D image. For instance, edge locations are surfaces in 3D cases which would be much more challenging to handle compared to edge curves in 2D cases. We propose a novel 3D image denoising procedure in this paper, based on local approximation of the edge surfaces using a set of surface templates. An important property of this method is that it can preserve edges and major edge structures (e.g., intersections of two edge surfaces and pointed corners). Numerical studies show that it works well in various applications.     

Image steganography using deep learning based edge detection

This paper introduces a deep learning-based Steganography method for hiding secret information within the cover image. For this, we use a convolutional neural network (CNN) with Deep Supervision based edge detector, which can retain more edge pixels over conventional edge detection algorithms. Initially, the cover image is pre-processed by masking the last 5-bits of each pixel. The said edge detector model is then applied to obtain a gray-scale edge map. To get the prominent edge information, the gray-scale edge map is converted into a binary version using both global and adaptive binarization schemes. The purpose of using different binarization techniques is to prove the less sensitive nature of the edge detection method to the thresholding approaches. Our rule for embedding secret bits within the cover image is as follows: more bits into the edge pixels while fewer bits into the non-edge pixels. Experimental outcomes on various standard images confirm that compared to state-of-the-art methods, the proposed method achieves a higher payload.

     

Moving object detection using modified temporal differencing and local fuzzy thresholding

Most of the existing video object detection schemes are either computationally extensive or fail to detect moving objects in different challenging situations. In this paper, we propose a robust and computationally inexpensive scheme to detect moving objects in video. The threefold approach begins with computation of difference images using temporal information. Difference images are calculated by subtracting two input frames, at each pixel position. Instead of generating difference images using the traditional continuous frame difference approach, we propose using a fixed number of alternate frames centered around the current frame. This approach aids in reducing the computational complexity without compromising on quality of the difference images. After computation of difference images, a novel post-processing scheme is employed by utilizing gamma correction factor and Mahalanobis distance metric to reduce false positives and false negatives. Object segmentation is finally performed on the refined difference image by a local fuzzy thresholding scheme. This avoids problems that are usually encountered in hard thresholding, especially pixel misclassification, which is the most important one. For robust experimental analysis, videos from changedetction.net, CAVIAR, and http://perception.i2r datasets have been used. These selected videos contain a wide variety of common challenges faced during object detection. Some examples are the presence of dynamic backgrounds, shadows, bad weather, etc. The results establish the effectiveness of the proposed scheme over some of the existing schemes both qualitatively and quantitatively as delineated in the experimental result section.     

Robust analysis for improvement of vowel onset point detection under noisy conditions

Vowel onset point (VOP) is the instant of time at which the vowel region starts in a speech signal. The VOPs are used as anchor points to design various speech based systems. Different algorithms exist in the literature to identify the occurrences of vowels in continuous spoken utterances. The algorithm based on combined evidences derived from source excitation, spectral peaks and modulation spectrum have been used as a baseline system for the present study. The baseline system provides a satisfactory level of performance under clean data condition. However under noisy data condition the performance of the previous system may be improved further by additional pre-processing of the raw speech data and post-processing the detected VOPs. In this paper we propose to use the speech enhancement techniques as pre-processing module to remove the noise from the speech data under different noisy conditions. The pre-processed speech data is then passed through the baseline system to detect the VOPs. It has been observed that there exist several spurious VOPs at the output of the baseline system. We propose to use a post-processing module based on average signal-to-noise ratio and information derived from the glottal closure instant to remove the spurious VOPs. The experiments were carried out on clean, artificially injected noisy, and data collected from the practical noisy environments. The results suggest that the proposed system using pre-processing and post-processing modules is robust and shows an improvement of 28–35 % over the existing baseline system by removing the spurious VOPs under different noisy conditions.