Unprocessed Data Research Articles

Automated Human Action Recognition with Improved Graph Convolutional Network-based Pose Estimation

The process of utilizing Artificial Intelligence (AI) to identify and label human behaviors from unprocessed activity data gathered from various sources is known as Human Activity Recognition (HAR). Because of its potential applications across multiple areas, computer vision faces a significant challenge in recognizing human actions and the accompanying interactions with objects and the environment. Investigating the temporal and geographical characteristics of the skeleton sequence is essential for this endeavor, according to recent studies. However, efficiently extracting discriminative temporal and spatial information remains a difficult task. This work proposes a novel Human Action Recognition Model exploiting improved Graph Convolutional Network (GCN)-based pose estimation with a Hybrid Classifier (IGCN-HC). The phases carried out in this model are pre-processing, pose estimation, feature extraction, and activity recognition. Initially, the input video will be pre-processed and a frame from the input video stream will be generated. Subsequently, human pose estimation exploiting improved GCN will be accomplished. Further, human skeletal joints’ coordinates in two- or three-dimensional spaces are determined via human pose estimation. Then, Shape Local Binary Texture (SLBT) and an improved hierarchy of skeleton features have been used to detect the variance in different activities. In the last phase, a hybrid classification model with the combination of Deep Maxout and customized CNN has been proposed for the recognition phase. The model utilizes two inputs pose estimation results (skeleton) and the extracted features for training purposes. Finally, the proposed trained model is evaluated for recognition on different test inputs and contrasted with the existing techniques.

Gearbox fault diagnosis: A DL-based nonparametric filter approach

Deep Learning (DL) models have become increasingly popular for gearbox fault diagnosis due to their ability to automatically extract features and deliver accurate results with time series data. Typically, these models are applied directly to raw data collected from industrial machines. However, such real-world data often contains significant noise and interference from other components, which can compromise signal quality and reduce the classification accuracy of DL models. This work investigates the effectiveness of signal preprocessing using the Energy Operator (EO) and its variants, non-parametric filters, to enhance the performance of convolutional neural networks (CNNs), a widely used deep learning technique, in fault diagnosis. By improving the signal-to-noise ratio (SNR), EO and its variants help the CNN classifier achieve better accuracy compared to when raw, unprocessed data is used. The findings highlight that the higher-order variant, EO123, significantly outperforms other EO variants, as well as the traditional approach of using raw data without preprocessing. This superior performance of EO123 is consistently demonstrated across three benchmark datasets, each representing different gearboxes under various fault conditions. The results highlight that EO123 is the most effective method for improving classification accuracy in DL-based gearbox fault diagnosis, making it the preferred choice over other preprocessing techniques and raw data.

An enhanced predictive modelling framework for highly accurate non-alcoholic fatty liver disease forecasting

Non-alcoholic fatty liver disease (NAFLD) is a chronic medical ailment characterized by accumulation of excessive fat in the liver of non-alcoholic patients. In absence of any early visible indications, application of machine learning based predictive techniques for early prediction of NAFLD are quite beneficial. The objective of this paper is to present a complete framework for guided development of varied predictive machine learning models and predict NAFLD disease with high accuracy. The framework employs step–by-step data quality enhancement to medical data such as cleaning, normalization, data upscaling using SMOTE (for handling class imbalances) and correlation analysis-based feature selection to predict NAFLD with high accuracy using only clinically recorded identifiers. Comprehensive comparative analysis of prediction results of seven machine learning predictive models is done using unprocessed as well as quality enhanced data. As per the observed results, XGBoost, random forest and neural network machine learning models reported significantly higher accuracies with improved ‘AUC’ and ‘ROC’ values using preprocessed data in contrast to unprocessed data. The prediction results are also assessed on various quality metrics such as ‘accuracy’, ‘f1-score’, ‘precision’, and ‘recall’ significantly support the need for presented methodologies for qualitative NAFLD prediction modelling.

Path Planning under High-dimensional Input States Based on Deep Q-Network

The field of autonomous navigation continues to face challenges in path planning, particularly when addressing the complex, high-dimensional input states that conventional algorithms struggle to process efficiently. This study presents a novel path-planning approach that utilizes Deep Q-Networks (DQN) to manage intricate and multidimensional environmental data. By integrating a DQN with path planning, this study aims to develop an adaptive system capable of making real-time decisions in dynamic environments. The methodology involves training a neural network to approximate the Q-function, enabling the agent to learn optimal strategies directly from unprocessed sensor data such as visual or LiDAR inputs. Experiments conducted in both simulated and real-world scenarios demonstrated the efficacy of this method, revealing significant improvements in route optimization, computational efficiency, and robustness against unforeseen obstacles compared to traditional techniques. The proposed system was evaluated in diverse settings, including urban environments and challenging terrain, illustrating its versatility. These findings suggest that DQN-based path planning has considerable potential for applications in robotics, autonomous vehicles, and other domains requiring intelligent decision-making under uncertainty.

Aberration correction-impact on image quality and chamber quantification in transthoracic echocardiography.

To improve image quality (IQ) in echocardiography, an aberration correction (AC) algorithm has recently been implemented in commercial scanners. We aimed to study (i) the correlation of a subjective IQ-score and an objective IQ-metric [global image coherence (GIC)], (ii) if AC improved IQ; (iii) if AC affected average values and interobserver agreement of left ventricular (LV) size, LV longitudinal strain, and left atrial (LA) volume. From 50 adult patients, where 45 (90%) had cardiovascular disease, unprocessed image data (channel data) were acquired from six standard transthoracic views. The data were processed with and without AC, resulting in 300 pairs of cine-loops. The cine-loops were randomly presented one-by-one to two blinded raters experienced in echocardiography. Both raters scored IQ subjectively from 1 (poor) to 4 (very good) and quantified LV dimensions, volumes and longitudinal strain, and LA volume. IQ-score correlated with GIC, Spearman rho 0.72, P < 0.001. AC improved median IQ-score from 2.5 to 3.0 (Wilcoxon signed rank: P < 0.001). The differences in average values of LV size, LV longitudinal strain, or LA volume with and without AC were not statistically significant and numerically minimal. Measured by intraclass correlation, interobserver agreement of these values was not significantly affected by AC. Image quality-score strongly correlated with GIC. Aberration correction improved IQ. However, AC did not lead to statistically significant changes in average values or interobserver agreement of LV size, LV longitudinal strain or LA volume quantification. Likely, the major benefit of AC is enhanced visualization of anatomical details.

Emotional privacy-preserving of speech based on generative adversarial networks

Consumer electronic devices with voice assistants are becoming increasingly popular in modern intelligent home. Nevertheless, directly uploading unprocessed speech data, which may contain sensitive attributes, to a cloud server poses a significant risk to user privacy. To address this privacy issue, this paper proposes a privacy-enhancing model to protect speech emotions based on generative adversarial networks (PSEGAN). The model aims to prevent the inference of emotional attributes while maintaining the accuracy and utility of speech features. PSEGAN benefits from three modules: (1) A pre-trained speaker matcher imposes generative constraints on the model during the training phase to ensure that the generated speech retains the essential information needed for speaker recognition. (2) Attribute adversarial networks can generate perturbed speech that transforms emotional attributes while preserving the utility of the speech. (3) Gated Recurrent Networks (GRN) can handle the long-short term dependencies of speech signals. PSEGAN model solves the problem of utility loss in traditional speech privacy preservation methods based on generative adversarial networks (GAN). Experimental results show that on the RAVDESS dataset, PSEGAN reduces emotion recognition accuracy by 80.7%, while speaker recognition accuracy only decreases by 1.1%. These findings demonstrate that PSEGAN effectively mitigates the leakage of emotional attributes while maintaining high utility.

Complexity, interpretability and robustness of GP-based feature engineering in remote sensing

Complexity, interpretability and robustness of GP-based feature engineering in remote sensing

Integration of FTIR Spectroscopy, Volatile Compound Profiling, and Chemometric Techniques for Advanced Geographical and Varietal Analysis of Moroccan Eucalyptus Essential Oils.

Eucalyptus essential oil is widely valued for its therapeutic properties and extensive commercial applications, with its chemical composition significantly influenced by species variety, geographical origin, and environmental conditions. This study aims to develop a reliable method for identifying the geographical origin and variety of eucalyptus oil samples through the application of advanced analytical techniques combined with chemometric methods. Essential oils from Eucalyptus globulus and Eucalyptus camaldulensis were analyzed using Gas Chromatography-Flame Ionization Detection (GC-FID) and Fourier Transform Infrared (FTIR) Spectroscopy. Chemometric analyses, including Orthogonal Partial Least Squares-Discriminant Analysis (O2PLS-DA) and Hierarchical Cluster Analysis (HCA), were utilized to classify the oils based on their volatile compound profiles. Notably, O2PLS-DA was applied directly to the raw FTIR data without additional spectral processing, showcasing its robustness in handling unprocessed data. For geographical origin determination, the GC-FID model achieved a Correct Classification Rate (CCR) of 100%, with 100% specificity and 100% sensitivity for both calibration and validation sets. FTIR spectroscopy achieved a CCR of 100%, specificity of 100%, and sensitivity of 100% for the calibration set, while the validation set yielded a CCR of 95.83%, specificity of 99.02%, and sensitivity of 94.44%. In contrast, the analysis based on species variety demonstrated 100% accuracy across all metrics CCR, specificity, and sensitivity-for both calibration and validation using both techniques. These findings underscore the effectiveness of volatile and infrared spectroscopy profiling for quality control and authentication, providing robust tools for ensuring the consistency and reliability of eucalyptus essential oils in various industrial and therapeutic applications.

ACTF: An efficient lossless compression algorithm for time series floating point data

ACTF: An efficient lossless compression algorithm for time series floating point data

Long-term load forecasting for smart grid

Abstract The load forecasting problem is a complicated non-linear problem connected with the weather, economy, and other complex factors. For electrical power systems, long-term load forecasting provides valuable information for scheduling maintenance, evaluating adequacy, and managing limited energy supplies. A future generating, transmission, and distribution facility's development and planning process begins with long-term demand forecasting. The development of advanced metering infrastructure (AMI) has greatly expanded the amount of real-time data collection on large-scale electricity consumption. The load forecasting techniques have changed significantly as a result of the real-time utilization of this vast amount of smart meter data. This study suggests numerous approaches for long-term load forecasting using smart-metered data from an actual distribution system on the NIT Patna campus. Data pre-processing is the process of converting unprocessed data into a suitable format by eliminating possible errors caused by lost or interrupted communications, the presence of noise or outliers, duplicate or incorrect data, etc. The load forecasting model is trained using historical load data and significant climatic variables discovered through correlation analysis. With a minimum MAPE and RMSE for every testing scenario, the proposed artificial neural network model yields the greatest forecasting performance for the used system data. The efficacy of the proposed technique has been through a comparison of the acquired results with various alternative load forecasting methods.

Characterization and Mitigation of a Simultaneous Multi-Slice fMRI Artifact: Multiband Artifact Regression inSimultaneous Slices.

Simultaneous multi-slice (multiband) acceleration in fMRI has become widespread, but may be affected by novel forms of signal artifact. Here, we demonstrate a previously unreported artifact manifesting as a shared signal between simultaneously acquired slices in all resting-state and task-based multiband fMRI datasets we investigated, including publicly available consortium data from the Human Connectome Project (HCP) and Adolescent Brain Cognitive Development (ABCD) Study. We propose Multiband Artifact Regression in Simultaneous Slices (MARSS), a regression-based detection and correction technique that successfully mitigates this shared signal in unprocessed data. We demonstrate that the signal isolated by MARSS correction is likely nonneural, appearing stronger in neurovasculature than gray matter. Additionally, we evaluate MARSS both against and in tandem with sICA+FIX denoising, which is implemented in HCP resting-state data, to show that MARSS mitigates residual artifact signal that is not modeled by sICA+FIX. MARSS correction leads to study-wide increases in signal-to-noise ratio, decreases in cortical coefficient of variation, and mitigation of systematic artefactual spatial patterns in participant-level task betas. Finally, MARSS correction has substantive effects on second-level t-statistics in analyses of task-evoked activation. We recommend that investigators apply MARSS to multiband fMRI datasets with moderate or higher acceleration factors, in combination with established denoising methods.

Quality assessment and control of unprocessed anatomical, functional, and diffusion MRI of the human brain using MRIQC.

Quality control of MRI data prior to preprocessing is fundamental, as substandard data are known to increase variability spuriously. Currently, no automated or manual method reliably identifies subpar images, given pre-specified exclusion criteria. In this work, we propose a protocol describing how to carry out the visual assessment of T1-weighted, T2-weighted, functional, and diffusion MRI scans of the human brain with the visual reports generated by MRIQC. The protocol describes how to execute the software on all the images of the input dataset using typical research settings (i.e., a high-performance computing cluster). We then describe how to screen the visual reports generated with MRIQC to identify artifacts and potential quality issues and annotate the latter with the "rating widget" - a utility that enables rapid annotation and minimizes bookkeeping errors. Integrating proper quality control checks on the unprocessed data is fundamental to producing reliable statistical results and crucial to identifying faults in the scanning settings, preempting the acquisition of large datasets with persistent artifacts that should have been addressed as they emerged.

Novel Genetic Optimization Techniques for Accurate Social Media Data Summarization and Classification Using Deep Learning Models

In the era of big data, effectively processing and understanding the vast quantities of brief texts on social media platforms like Twitter (X) is a significant challenge. This paper introduces a novel approach to automatic text summarization aimed at improving accuracy while minimizing redundancy. The proposed method involves a two-step process: first, feature extraction using term frequency–inverse document frequency (TF–IDF), and second, summary extraction through genetic optimized fully connected convolutional neural networks (GO-FC-CNNs). The approach was evaluated on datasets from the Kaggle collection, focusing on topics like FIFA, farmer demonstrations, and COVID-19, demonstrating its versatility across different domains. Preprocessing steps such as tokenization, stemming, stop word s removal, and keyword identification were employed to handle unprocessed data. The integration of genetic optimization into the neural network significantly improved performance compared to traditional methods. Evaluation using the ROUGE criteria showed that the proposed method achieved higher accuracy (98.00%), precision (98.30%), recall (98.72%), and F1-score (98.61%) than existing approaches. These findings suggest that this method can help create a reliable and effective system for large-scale social media data processing, enhancing data dissemination and decision-making.

Assessing the effects of artifacts and noise in EEG signals on car-following driving behavior prediction

Assessing the effects of artifacts and noise in EEG signals on car-following driving behavior prediction

Privacy-Preserving Computing Services for Encrypted Personal Data Through Streams Over Distributed Ledgers

The growing adoption of wearables is driving the demand for personalized services that leverage unprocessed data, such as biometric and health information, to enhance user experiences and support through software applications. However, several existing use cases involving this information still prioritize traditional schemes, neglecting user privacy. Consequently, the transparency of data transmission paths and the potential for tampering remain ambiguous when users share data with service providers. In this paper, we propose the application of an Internet of Things device-focused distributed ledger as an underlying layer for the transmission of encrypted data using streams. Moreover, our proposal enables data recording for future events and the implementation of multi-subscriber models, allowing client information to be shared securely with different service providers. Through simulation experiments conducted on constrained devices, we demonstrate that our proposed framework efficiently transmits large ciphertexts through streams on a distributed ledger, overcoming the inherent limitations of such networks when dealing with substantial data volumes. Ultimately, the performance metrics presented prove that the proposed model is suitable for real-world applications requiring continuous data collection by wearables and subsequent transmission to service providers.

Intrusion Event Classification of a Drainage Tunnel Based on Principal Component Analysis and Neural Networking

Drainage tunnel stability is crucial for engineering project safety (e.g., mine engineering and dams), and rockfall events and water release are key indicators of drainage tunnel stability. To address this, we developed a monitoring system to simulate drainage tunnel intrusions based on distributed acoustic sensing (DAS), and we obtained typical characteristics of events like rockfall events and water release. Given the multitude of DAS signal feature parameters and challenges, such as high-dimensional features impacting the classification accuracy of machine learning, we proposed an identification method for drainage tunnel intrusion events using principal component analysis (PCA) and neural networks. PCA reveals that amplitude-related parameters—amplitude, mean amplitude, and energy—significantly contribute to DAS signal classification, reducing the feature parameter dimensions by 54.8%. The accuracy of intrusion event classification improves with PCA-processed data compared to unprocessed data, with overall accuracy rates of 79.1% for rockfall events and 72.7% for water release events. Additionally, the artificial neural network model outperforms the Bayesian and logistic regression models, demonstrating that ANN has advantages in handling complex models for intrusion event classification.

The Human Connectome Project of adolescent anxiety and depression dataset

This article describes primary data and resources available from the Boston Adolescent Neuroimaging of Depression and Anxiety (BANDA) study, a novel arm of the Human Connectome Project (HCP). Data were collected from 215 adolescents (14–17 years old), 152 of whom had current diagnoses of anxiety and/or depressive disorders at study intake. Data include cross-sectional structural (T1- and T2-weighted), functional (resting state and three tasks), and diffusion-weighted magnetic resonance images. Both unprocessed and HCP minimally-preprocessed imaging data are available within the data release packages. Adolescent and parent clinical interview data, as well as cognitive and neuropsychological data are also included within these packages. Release packages additionally provide data collected from self-report measures assessing key features of adolescent psychopathology, including: anxious and depressive symptom dimensions, behavioral inhibition/activation, exposure to stressful life events, and risk behaviors. Finally, the release packages include 6- and 12-month longitudinal data acquired from clinical measures. Data are publicly accessible through the National Institute of Mental Health Data Archive (ID: #2505).

Characterization of facial and ocular gestures through electroencephalogram

This article describes the characterization of facial and ocular gestures using the electroencephalogram (EEG) method connected with an EMOTIV EPOC+ Brainwear® device. This characterization is developed by the storage of raw data (unprocessed data) acquired by the device. The experiment was applied to nine subjects, considering that EEG explores neurophysiologically with high levels of statistical confidence the bioelectric activity in the brain in the condition of resting state such as wakeups or dreaming states. In contrast to non-resting states, the registered data showed a random and distinct activation of hyperpnea and intermittent luminous stimulus. Despite the reduced number of samples in the experiment, the results showed that the level of confidence was greater than 75%. The data was characterized and processed by a support vector machine (SVM).

Quantification of Two-Dimensional Interfaces: Quality of Heterostructures and What Is Inside a Nanobubble.

Trapped materials at the interfaces of two-dimensional heterostructures (HS) lead to reduced coupling between the layers, resulting in degraded optoelectronic performance and device variability. Further, nanobubbles can form at the interface during transfer or after annealing. The question of what is inside a nanobubble, i.e., the trapped material, remains unanswered, limiting the studies and applications of these nanobubble systems. In this work, we report two key advances. First, we quantify the interface quality using RAW format optical imaging (unprocessed image data) and distinguish between ideal and non-ideal interfaces. The HS/substrate ratio value is calculated using a transfer matrix model and is able to detect the presence of trapped layers. The second key advance is the identification of water as the trapped material inside a nanobubble. To the best of our knowledge, this is the first study to show that optical imaging alone can quantify interface quality and find the type of trapped material inside spontaneously formed nanobubbles. We also define a quality index parameter to quantify the interface quality of HS. Quantitative measurement of the interface will help answer the question whether annealing is necessary during HS preparation and will enable creation of complex HS with small twist angles. Identification of the trapped materials will pave the way toward using nanobubbles for optical and engineering applications.

Enhancing Predictive Modeling with Human-Like Intelligence: A Deep Feature Synthesis Approach

In this work, we create the Data Science Machine, an automated tool for extracting predictive models from unprocessed data. We initially propose and build the Deep Feature Synthesis method for automatically creating features for relational datasets in order to accomplish this automation. The pursuit of Human-Like Intelligence (HLI) in AI systems, the creation of emotionally intelligent AI, and the possible convergence of XAI with cognitive sciences are all further explored in this study. The advancement of artificial intelligence (AI) towards Artificial General Intelligence (AGI) raises important questions about consciousness, ethics, and social consequences.