Search

With the development and advent of Internet communications and multimedia technology, it is of high importance to secure the digital media, including audio, video, speech, and images. In fact, watermarking technique provides an efficient solution to this problem. However, for an effective watermarking of data, several important requirements need to be met, such as transparency, low computational cost, robustness, and security. In this work, a robust hybridized medical images watermarking approach based on Walsh Hadamard Transform (WHT) and singular value decomposition (S VD) is presented. Both of the host image and the watermark images were modified with singular values obtained by the (S VD) approach. The simulation results demonstrate that this proposed watermarking framework is highly robust and achieves high peak signal to noise ratio (PS NR) and high structure similarity index (S S IM). [ABSTRACT FROM AUTHOR]

Estimating the number of latent topics is a prerequisite to using topic modeling algorithms such as LDA. A number of methods for this purpose are compared, including the de facto approach based on perplexity. Results from synthetic data show that a combination of methods yields better estimates than any single method, and in particular it provides an indicator of the reliability of the estimated number of topics, something no single method can do. In this paper, we extend the findings based on synthetic data over real data and introduce a technique, "Cosine inter-intra topics", to assess the validity of the ground truth of real data. Moreover, we measure the reliability of estimated number of topics through the variance of methods. Results corroborate the ones from synthetic data, suggesting the approach generalizes to real text corpus. [ABSTRACT FROM AUTHOR]

The classification of pathological diseases with the implementation of concepts of Deep Learning has been increasing considerably in recent times. Among the works developed there are good results for the classification in sustained speech with vowels, but few related works for the classification in continuous speech. This work uses the German Saarbrücken Voice Database with the phrase " Guten Morgen, wie geht es Ihnen?" to classify four classes: dysphonia, laryngitis, paralysis of vocal cords and healthy voices. Transfer learning concepts were used with the AudioSet database. Two models were developed based on Long-Short-Term-Memory and Convolutional Network for classification of extracted embeddings and comparison of the best results, using cross-validation. The final results allowed to obtaining 40% of f1-score for the four classes, 66% f1-score for Dysphonia x Healthy, 67% for Laryngitis x healthy and 80% for Paralysis x Healthy. [ABSTRACT FROM AUTHOR]

Research in the fields of machine learning and intelligent systems addresses essential problem of developing computer algorithms that can deal with huge amounts of data and then utilize this data in an intellectual way to solve a variety of real-world problems. In many applications, to interpret data with a large number of variables in a meaningful way, it is essential to reduce the number of variables and interpret linear combinations of the data. Principal Component Analysis (PCA) is an unsupervised learning technique that uses sophisticated mathematical principles to reduce the dimensionality of large datasets. The goal of this paper is to provide a complete understanding of the sophisticated PCA in the fields of machine learning and data dimensional reduction. It explains its mathematical aspect and describes its relationship with Singular Value Decomposition (SVD) when PCA is calculated using the covariance matrix. In addition, with the use of MATLAB, the paper shows the usefulness of PCA in representing and visualizing Iris dataset using a smaller number of variables. [ABSTRACT FROM AUTHOR]

Abstract Scene classification is considered as an imperative issue for computer vision and has got extensive consideration in the recent past. Due to recent developments in high performance computing units such as GPUs, popularly known deep learning algorithm namely, convolutional neural networks (CNNs), exploits huge datasets to give powerful models. The paper proposes the use of transfer learning technique, by which a pre-trained model known as Places-CNN is used to generate feature vectors for each scene image of the dataset. The scene-classification experiments are conducted on the Oliva Torralba (OT) scene dataset, which consists of eight outdoor scene categories. The features were extracted from the fully connected layer of the pre-trained Places CNN architecture. The deep features were extracted from the input color images and the grayscale images converted using two different techniques based on singular value decomposition (SVD). The results obtained from classification experiments show that, models trained on SVD-Decolorized and Modified-SVD decolorized images give comparable performance to the input color images. Unlike the color images, which use three planes (RGB) of information, the grayscale images use only one plane of information. The grayscale images were able to retain the required shape and texture information from the original RGB images and, thus sufficient to categorize the classes of scene images. [ABSTRACT FROM AUTHOR]

Fetal heart monitoring yields vital information about the fetus health and can support medical decision making in critical situations. A compound signal is obtained noninvasively by placing electrodes on the abdomen area of the mother which contains maternal and fetal ECG signals contaminated by various other signals from body and externally induced noises. As a result, the basic problem is to extract the fECG signal from the mixture of mECG and fECG, where the interfering mECG is a much stronger. This problem has been addressed in several works, in this paper, we propose a novel blind-source separation method to extract fetal ECG when given only a single channel recording. Our approach is based on the wavelet theory and SVD-ICA methods. It is pointed out that the presented algorithm is validated by Matlab simulation and can be implemented on real time life by using embedded system such as DSP processor. [ABSTRACT FROM AUTHOR]

The paper presents several thresholds obtained by heuristic approach for face verification using Locally Adaptive Regression Kernel (LARK) descriptors for euclidean, cosine and chebyshev distance metrics. The absence of a threshold for several distance metrics possess several setbacks such as increased computational complexity and escalated runtime. The proposed method has significantly higher influence when the verification process is computationally intensive. The proposed approach requires minimal training and face verification is accomplished based on the threshold value obtained during training. The whole process is modest and nearly all of the time spent is solely on quantifying any of the distance metrics between the LARKs obtained from the two faces for verification. LARK descriptors compute a measure of resemblance on the basis of “signal-induced distance” between a pixel and its nearby pixels. We assess the interspace between the LARKs from these faces and analogize the resemblance from the threshold resulting in a binary decision. [ABSTRACT FROM AUTHOR]

The rapid increase in usage of personal computers, internet and digital multimedia technology leads to easily sharing of digital data/ media. However, availability of numerous image processing tools facilitate unauthorized use of such data. Unauthorized users/ attackers can easily copy, delete or modify digital data. This problem of illegal modification/ reproduction of digital data, leads to innovation of some techniques which can protect intellectual property rights of digital data/ media. Recently, watermarking has been identified as a major tool to attain copyright protection/ authentication. A digital watermark can be embedded in host data in spatial domain as well as in frequency domain. In this work a hybridized technique incorporating Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD) has been presented. [ABSTRACT FROM AUTHOR]

In the process of medical image sharing, it is always required that while sending the medical images, the images must be secured so that it does not get tampered or attacked and the integrity of the image and watermark are maintained. This paper presents a technique for authentication of medical image using zero watermarking method. The image is applied with Discrete Wavelet Transform (DWT) method which divides the image into four bands. After applying the DWT, blocking is done and each block is treated with SVD to extract unique feature and a master share is obtained by performing logical operation between extracted unique feature and the watermark. The strategy gives better experimental results and can be utilized in the field of telemedicine in order to protect the shared images from any unlawful use. [ABSTRACT FROM AUTHOR]

The growth of the Internet along with the increasing availability of multimedia applications has given rise to the number of issues in medical science and defense applications. The ease of communication has given rise to several privacy problems which can be addressed by watermarking. This paper presents a robust watermarking technique based on texture matching of the watermark with the host image. The watermark is texturized using Arnold transform and matching is found using histogram of gradient and Log Gabor filter. The watermark is inserted at several places in the image to ensure the security of the image in smaller parts. The efficacy of the algorithm is tested in terms of PSNR (peak signal to noise ratio) and NCC (normalized correlation coefficient). [ABSTRACT FROM AUTHOR]

In the gargantuan domain of biometrics, the most prominent field is face recognition. We are our faces, in a way we are not our social networking profiles, our legal names and Aadhaar identification number. Even though voluminous data are collected online about most individuals, for instance on web using browser cookies, ip addresses, MAC addresses or email addresses. Almost everything that represents an individual is merely an untidy collection or pile of numbers and letters. All of these can be changed with some cost or sacrifice. Today, we hear that the victims of fraud can apply for obtaining new unique identifier(s). Despite these, there remains one unique identifier that's different from these and that is our face. It's arduous to change it beyond recognition, if it's even feasible. That is, face recognition bind data about us to us only. Thus, face recognition aids law enforcement agencies as a crime-fighting tool to recognize people based on facial traits. The recent stoor of this field has shown its importance in real time applications. This has created an exponential impact on the research work being carried out in this field over the last few decades. In the recent past binary descriptor based techniques like SIFT, SURF, etc are being widely deployed for recognition systems. Keeping this as focal point, the paper proposes two dimensionality reduction techniques namely SVD (Singular Value Decomposition) and PCA (Principal Component Analysis) for SURF based face recognition. The results of simulations conducted on four exemplar datasets show that the SURF-SVD method is more efficient for face recognition when compared with the other existing methods including the SURF-PCA method. [ABSTRACT FROM AUTHOR]

Current recommender systems are very inefficient. There are many metrics that are used to measure the effectiveness of recommender systems. These metrics often include “conversion rate” and “click through rate”. Recently, these rates are in low single digit (less than 10%). In other words, more than 90% of times, the model that the targeting system is based on, produces noise. The belief in this work is that the main problem leading to getting such unsatisfactory outcomes is the modeling problem. Much of the modeling problem could be represented and exemplified in treating users and items as member of clusters (segments). In this work, we consider full personalization of recommendation systems. We aim at personalization of users and items simultaneously. [ABSTRACT FROM AUTHOR]

This paper describes the comparative study of Truncated DCT-SVD and DWT-SVD. In this paper we propose two different approaches to compute the feature vector for content based image retrieval (CBIR) system. SVD feature of successively truncated DCT image and DWT decomposed image computed for grayscale image, RGB and YCbCrcolor image. Truncated DCT and DWT decomposition SVD features of the image computed up to fifth level to compare the performance. Proposed methods incorporate with the multidimensional features vector computed by using SVD of low frequency coefficients of DCT and DWT of image. Similarity between the query image and database image measured here by using simple Euclidean distance and Bray Curtis Distance. The overall average precision and average recall crossover point of each image category. Proposed methods are tested on the augmented image database. [ABSTRACT FROM AUTHOR]

Image denoising without any idea about kind of noise is very cumbersome as compared to known noise. This paper proposed a novel technique for blind image denoising using SVD and local pixel grouping. The technique is checked against salt & pepper noise and Gaussian noise for gray as well as color images and compared with state of art algorithm LPGPCA. It is found that the proposed technique gives better results when compared using objective criteria. [ABSTRACT FROM AUTHOR]