International Journal of Information Technology最新文献

Visual phrase recognition needs lip movement related visual features, while audio-visual phrase recognition requires both acoustic and visual features. In this work, we propose a novel visual feature; Spatio-temporal Weber Gradient Directional (SWGD) to effectively represent the micro-patterns of lip movements. The proposed visual feature is obtained by using micro-texture information; local differential excitation, gradient orientation, and gradient directional information. Experiments are conducted using standard OuluVS database. Polynomial kernel based support vector machine (SVM) classifier is employed, as it provides relatively better performance. The SWGD extracted from (2times 5times 3) video block size provides higher performance of 73.9%. Additionally, we explore twelve distinct local descriptors commonly employed in face recognition and utilize them for the first time in a comparative study of phrase recognition. SWGD performs better than these twelve distinct features but has higher dimension of 4320. By reducing the dimension to 100 using the soft locality preserving map (SLPM), performance improved from 73.9 to 81.3%. The dimensionally reduced SWGD (SWGD(_{text {SLPM}})) outperforms other state-of-the-art visual features mentioned in this paper. This shows the benefit of the salient micro-texture information considered in the proposed feature but neglected in state-of-the-art features. We observe that the SWGD(_{text {SLPM}}) feature has high discriminative ability to represent distinct lip movement patterns for different phrases. Mel-frequency cepstral coefficient (MFCC) based audio phrase recognizer performance degrades as the signal-to-noise level decreases. Including the SWGD(_{text {SLPM}}) visual feature and Glottal MFCC (GMFCC) excitation source feature improves performance by 3.6%, reflecting noise robustness.

This study presents the Normal Discriminant Feature Selection based Regressive Deep Neural MapReduce (NDFS-RDNMR) framework designed for efficient prediction of diabetic chronic diseases using input datasets. The primary aim of NDFS-RDNMR is to enhance accuracy and recall in handling large datasets for chronic disease prediction. The framework integrates the Normal Discriminative Preprocessing Model (NDPM) and bivariant regressive deep artificial neural network with MapReduce (BRDANNMR) classifier. Utilizing the Pima Indian diabetic dataset as input, NDFS-RDNMR conducts feature preprocessing through NDPM to extract relevant features for disease prediction. Non-traditional datasets are transformed into traditional formats via parameter rescaling to fit within predefined value ranges. Min–max normalization is applied to improve system accuracy while preserving data relationships. The BRDANNMR classifier utilizes bivariant regression analysis in the mapping phase to generate intermediary outcomes, which are then classified using a bipolar activation function in the reducer process. The framework achieves high accuracy and recall in early diabetes disease prediction, offering valuable insights for medical practitioners and researchers.

Smishing is a type of social engineering attack that involves sending fraudulent SMS messages to trick recipients into revealing sensitive information. In recent years, it has become a significant threat to mobile communications. In this study, we introduce a novel smishing detection method based on federated learning, which is a decentralized approach ensuring data privacy. We develop a robust detection model within a federated learning framework based on deep learning methods such as Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM). Our experiments show that the federated learning method using Bi-LSTM achieves an accuracy of 88.78%, highlighting its effectiveness in tackling smishing detection while preserving user privacy. This approach not only offers a promising solution to smishing attacks but also lays the groundwork for future research in mobile security and privacy-preserving machine learning.

Image steganography is the technique of masking secret information inside a carrier medium without disclosing the presence of the secret. Image steganography considers images to be carriers. Pixel Value Differencing (PVD) steganography is a spatial domain image steganographic technique in which secret messages are transformed into a sequence of bits, which are then concealed in the difference value of pixel intensities. The evaluation of an image steganography system typically hinges upon three pivotal metrics: hiding capacity, imperceptibility, and robustness. In this study, we aim to enhance the payload capacity of the system by modifying the pixel differences between each pair of pixels using the entropy value of the cover image. The findings from our experiments suggest a marked enhancement in payload capacity compared to the traditional PVD process, while maintaining satisfactory visual quality.

A new approach has been devised to assess the reliability of non-series-parallel (NSP) system, employing intuitionistic fuzzy concept, particularly focusing on triangular intuitionistic fuzzy numbers (TIFNs), alongside the path tracing technique. The system encompasses seven distinct components categorized into three subsystems. Two parallel subsystems each consist of three components connected in series, while the third subsystem involves a single component linked with the extreme components of the parallel subsystems. The failure rates of the components are assumed as time-varying triangular intuitionistic fuzzy numbers. Subsequently, the reliability and Mean Time to System Failure (MTSF) expressions containing both membership and non-membership degrees have been derived utilizing path tracing and (α, β)-cut approach. This methodology is then applied to a Resistor-Inductor-Capacitor (RLC) system, and its intuitionistic fuzzy reliability and MTSF are evaluated. Graphical representations have been utilized to enhance comprehension of the reliability characteristics of the system.