Window Size Research Articles

Local-Peak Scale-Invariant Feature Transform for Fast and Random Image Stitching.

Image stitching aims to construct a wide field of view with high spatial resolution, which cannot be achieved in a single exposure. Typically, conventional image stitching techniques, other than deep learning, require complex computation and are thus computationally expensive, especially for stitching large raw images. In this study, inspired by the multiscale feature of fluid turbulence, we developed a fast feature point detection algorithm named local-peak scale-invariant feature transform (LP-SIFT), based on the multiscale local peaks and scale-invariant feature transform method. By combining LP-SIFT and RANSAC in image stitching, the stitching speed can be improved by orders compared with the original SIFT method. Benefiting from the adjustable size of the interrogation window, the LP-SIFT algorithm demonstrates comparable or even less stitching time than the other commonly used algorithms, while achieving comparable or even better stitching results. Nine large images (over 2600 × 1600 pixels), arranged randomly without prior knowledge, can be stitched within 158.94 s. The algorithm is highly practical for applications requiring a wide field of view in diverse application scenes, e.g., terrain mapping, biological analysis, and even criminal investigation.

Fog-assisted de-duplicated data exchange in distributed edge computing networks

The Internet of Things (IoT) generates substantial data through sensors for diverse applications, such as healthcare services. This article addresses the challenge of efficiently utilizing resources in resource-scarce IoT-enabled sensors to enhance data collection, transmission, and storage. Redundant data transmission from sensors covering overlapping areas incurs additional communication and storage costs. Existing schemes, namely Asymmetric Extremum (AE) and Rapid Asymmetric Maximum (RAM), employ fixed and variable-sized windows during chunking. However, these schemes face issues while selecting the index value to decide the variable window size, which may remain zero or very low, resulting in poor deduplication. This article resolves this issue in the proposed Controlled Cut-point Identification Algorithm (CCIA), designed to restrict the variable-sized window to a certain threshold. The index value for deciding the threshold will always be larger than the half size of the fixed window. It helps to find more duplicates, but the upper limit offset is also applied to avoid the unnecessarily large-sized window, which may cause extensive computation costs. The extensive simulations are performed by deploying Windows Communication Foundation services in the Azure cloud. The results demonstrate the superiority of CCIA in various metrics, including chunk number, average chunk size, minimum and maximum chunk number, variable chunking size, and probability of failure for cut point identification. In comparison to its competitors, RAM and AE, CCIA exhibits better performance across key parameters. Specifically, CCIA outperforms in total number of chunks (6.81%, 14.17%), average number of chunks (4.39%, 18.45%), and minimum chunk size (153%, 190%). These results highlight the effectiveness of CCIA in optimizing data transmission and storage within IoT systems, showcasing its potential for improved resource utilization and reduced operational costs.

Tuning the Transparency Window of SrVO3 Transparent Conducting Oxide.

Correlated transparent conducting oxides (TCOs) have gained great attention, because of their unique combination of transparency and metallic character. SrVO3 (SVO) was identified as a high-performance TCO in the visible range. Few studies have investigated band structure engineering through chemical doping to enhance the optical properties of SVO. Here, we use two different strategies by exploiting the band-filling and width of the bands derived from Vanadium to tune the screened plasma frequency ωp* and the interband transition Ep-d energy, corresponding to the optical transparency window edges. For control of the band-filling strategy, it is found that Titanium doped SVO has a wide transparency window, but such a composition does not maintain the high electrical conductivity required for TCO applications. Concerning the bandwidth strategy, the doping of SrVO3 by Calcium shows that ωp* remains located in the IR range (1.12 eV), while Ep-d is blue-shifted into the UV region (3.43 eV) due to reinforced electronic correlations. By an appropriate choice of dopant, we successfully increased the size of the transparency window by around 11% from 1.94 eV (SVO) to 2.30 eV (Calcium-doped SVO), while retaining high conductivity of around 2.30 × 104 (S·cm-1) and high charge carrier density of 2.93 × 1022 cm-3.

Total Magnetic Field Perturbations at Martian Low Altitudes (≤500 km): MAVEN Observations in 2014–2023

AbstractMagnetic perturbations in the Martian ionosphere are often considered as representing plasma irregularities, but thorough statistical comparisons between the two independent phenomena have been rarely conducted. In this study, we investigate the statistical relationship between total magnetic field perturbations and those of O2+ density as observed by the Mars Atmosphere and Volatile EvolutioN (MAVEN) spacecraft at altitudes below 500 km in 2014–2023. From the 1 Hz time series of magnetic field strength, perturbation intensity is estimated by subtracting a signal smoothed with a Savitzky‐Golay filter (window size ∼36 km). Statistically, the magnetic perturbations are stronger on the dayside than at night, in the local summer hemisphere than in winter, and away from crustal magnetic anomalies than nearby. Also, day‐to‐day variability of the magnetic perturbations exhibits a weak but significant correlation with that of solar wind electron density. The statistical behavior of total magnetic field perturbations is not the same as that of O2+ density perturbations: that is, the former is not a perfect proxy for the latter. The seemingly counter‐intuitive discrepancy can be explained by effects of inhomogeneous background magnetic field strength (participating in total pressure balance across plasma perturbations) and localized ionospheric currents at altitudes unreachable by MAVEN.

Dealing With Assumptions and Sampling Bias in the Estimation of Effective Population Size: A Case Study in an Amphibian Population.

Accurately estimating effective population size (N e) is essential for understanding evolutionary processes and guiding conservation efforts. This study investigates N e estimation methods in spatially structured populations using a population of moor frog (Rana arvalis) as a case study. We assessed the behaviour of N e estimates derived from the linkage disequilibrium (LD) method as we changed the spatial configuration of samples. Moor frog eggs were sampled from 25 breeding patches (i.e., separate vernal ponds, ditches or parts of larger fens) within a single population, revealing an isolation-by-distance pattern and a local spatial genetic structure. Varying buffer sizes around each patch were used to examine the impact of sampling window size on the estimation of effective number of breeders (N b). Our results indicate a downward bias in LD N b estimates with increasing buffer size, suggesting an underestimation of N b. The observed bias is attributed to LD resulting from including genetically divergent individuals (mixture-LD) confounding LD due to drift. This emphasises the significance of considering even subtle spatial genetic patterns. The implications of these findings are discussed, emphasising the need to account for spatial genetic structure to accurately assess population viability and inform conservation efforts. This study contributes to our understanding of the challenges associated with N e estimation in spatially structured populations and underscores the importance of refining methodologies to address population-specific spatial dynamics for effective conservation planning and management.

Combining optical flow and Swin Transformer for Space-Time video super-resolution

Space–time video super-resolution is a task that aims to interpolate low frame rate, low resolution videos to high frame rate, high resolution ones. While existing Transformer-based methods have achieved results comparable to convolutional neural networks-based methods, the computational cost of Transformer limits its performance with constrained computational resources. Moreover, Swin Transformer may fail to fully exploit the spatio-temporal information of video frames due to the limitation of window size, impeding its effectiveness in handling large motions. To address these limitations, we propose an end-to-end space–time video super-resolution architecture based on optical flow alignment and Swin Transformer. The alignment module is introduced to extract spatio-temporal information from adjacent frames without significantly increasing the computational burden. Additionally, we design a residual convolution layer to enhance the translational invariance of the features extracted by Swin Transformer and introduces additional nonlinear transformations. Experimental results demonstrate that our proposed method achieves superior performance on various benchmark datasets compared to state-of-the-art methods. In terms of Peak Signal-to-Noise Ratio, our method outperforms the state-of-the-art methods by at least 0.15 dB on Vimeo-Medium dataset.

Speed-Aware Audio-Driven Speech Animation using Adaptive Windows

We present a novel method that can generate realistic speech animations of a 3D face from audio using multiple adaptive windows. In contrast to previous studies that use a fixed size audio window, our method accepts an adaptive audio window as input, reflecting the audio speaking rate to use consistent phonemic information. Our system consists of three parts. First, the speaking rate is estimated from the input audio using a neural network trained in a self-supervised manner. Second, the appropriate window size that encloses the audio features is predicted adaptively based on the estimated speaking rate. Another key element lies in the use of multiple audio windows of different sizes as input to the animation generator: a small window to concentrate on detailed information and a large window to consider broad phonemic information near the center frame. Finally, the speech animation is generated from the multiple adaptive audio windows. Our method can generate realistic speech animations from in-the-wild audios at any speaking rate, i.e., fast raps, slow songs, as well as normal speech. We demonstrate via extensive quantitative and qualitative evaluations including a user study that our method outperforms state-of-the-art approaches.

The Effect of Window Size and Window Shape in STFT for Pre-Processing FMCW Radar Data in Human Activity Recognition Based on Bi-LSTM

The Effect of Window Size and Window Shape in STFT for Pre-Processing FMCW Radar Data in Human Activity Recognition Based on Bi-LSTM

Automated analysis of the void structure in hardened concrete based on shape from focus

The quality of the air void system is essential for the frost resistance of concrete. Therefore, it is crucial to assess the quality of the air void system in hardened concrete in an appropriate way. In this study, the shape from focus (SFF) method is applied to identify the focused image of each point on the polished surface of hardened concrete and thereby acquire the depth information of the entire surface. The performance of various focus measure operators and window sizes is evaluated. Thereafter, the selected focus measure operator and window size are applied for the SFF analysis of all samples. Based on the obtained depth map, the voids are identified, and the void parameters of the hardened concrete are determined by an automated procedure. The main advantage of the SFF method is that it is possible to carry out an automated air void analysis without contrast enhancement.

Deciphering the Language of Protein-DNA Interactions: A Deep Learning Approach Combining Contextual Embeddings and Multi-Scale Sequence Modeling

Deciphering the mechanisms governing protein-DNA interactions is crucial for understanding key cellular processes and disease pathways. In this work, we present a powerful deep learning approach that significantly advances the computational prediction of DNA-interacting residues from protein sequences.Our method leverages the rich contextual representations learned by pre-trained protein language models, such as ProtTrans, to capture intrinsic biochemical properties and sequence motifs indicative of DNA binding sites. We then integrate these contextual embeddings with a multi-window convolutional neural network architecture, which scans across the sequence at varying window sizes to effectively identify both local and global binding patterns.Comprehensive evaluation on curated benchmark datasets demonstrates the remarkable performance of our approach, achieving an area under the ROC curve (AUC) of 0.89 – a substantial improvement over previous state-of-the-art sequence-based predictors. This showcases the immense potential of pairing advanced representation learning and deep neural network designs for uncovering the complex syntax governing protein-DNA interactions directly from primary sequences.Our work not only provides a robust computational tool for characterizing DNA-binding mechanisms, but also highlights the transformative opportunities at the intersection of language modeling, deep learning, and protein sequence analysis. The publicly available code and data further facilitate broader adoption and continued development of these techniques for accelerating mechanistic insights into vital biological processes and disease pathways.In addition, the code and data for this work are available at https://github.com/B1607/DIRP.

ABBR: An augmented BBR for collaborative intelligent transmission over heterogeneous networks in IIoT

In the era of Industry 5.0, with the deep convergence of Industrial Internet of Things (IIoT) and 5G technology, stable transmission of massive data in heterogeneous networks becomes crucial. This is not only the key to improving the efficiency of human–machine collaboration, but also the basis for ensuring system continuity and reliability. The arrival of 5G has brought new challenges to the communication of IIoT in heterogeneous environments. Due to the inherent characteristics of wireless networks, such as random packet loss and network jitter, traditional transmission control schemes often fail to achieve optimal performance. In this paper we propose a novel transmission control algorithm, aBBR. It is an augmented algorithm based on BBRv3. aBBR dynamically adjusts the sending window size through real-time analysis to enhance the transmission performance in heterogeneous networks. Simulation results show that, compared to traditional algorithms, aBBR demonstrates the best comprehensive performance in terms of throughput, latency, and retransmission. When random packet loss exists in the link, aBBR improves the throughput by an average of 29.3% and decreases the retransmission rate by 18.5% while keeping the transmission delay at the same level as BBRv3.

Optimizing Windowing Techniques to Improve the Accuracy of Artificial Neural Networks in Predicting Outpatient Visits

Hospitals are healthcare institutions that play an important role in providing health services to the community. To optimize the service, hospitals need to predict the number of outpatient visits. The objectives of this research are (1) determine the effect of window size on the accuracy of predicting the number of outpatient visits, and (2) identify the best window size and accuracy of neural networks in predicting the daily number of outpatient visits. To achieve the research objectives, the following steps were undertaken: data collection of outpatient visits at RSUD dr. Soedirman Kebumen from 2018 to 2023, preprocessing, applying different window sizes, modeling neural networks, and testing by calculating the RMSE value for each window size. The test results show that the lowest RMSE for 2018 was 1.267 with a window size of 34, for 2019 was 1.262 with a window size of 34, for 2020 was 1.515 with a window size of 17, for 2021 was 1.81 with a window size of 18, for 2022 was 1.282 with a window size of 20, and for 2023 was 1.263 with a window size of 29. These window sizes indicate the cycle of outpatient visits each year. By understanding these visit cycles, the number of outpatient visits can be predicted at any time.

CTRNet: An Automatic Modulation Recognition Based on Transformer-CNN Neural Network

Deep learning (DL) has brought new perspectives and methods to automatic modulation recognition (AMR), enabling AMR systems to operate more efficiently and reliably in modern wireless communication environments through its powerful feature learning and complex pattern recognition capabilities. However, convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which are used for sequence recognition tasks, face two main challenges, respectively: the ineffective utilization of global information and slow processing speeds due to sequential operations. To address these issues, this paper introduces CTRNet, a novel automatic modulation recognition network that combines a CNN with Transformer. This combination leverages Transformer’s ability to adequately capture the long-distance dependencies between global sequences and its advantages in sequence modeling, along with the CNN’s capability to extract features from local feature regions of signals. During the data preprocessing stage, the original IQ-modulated signals undergo sliding-window processing. By selecting the appropriate window sizes and strides, multiple subsequences are formed, enabling the network to effectively handle complex modulation patterns. In the embedding module, token vectors are designed to integrate information from multiple samples within each window, enhancing the model’s understanding and modeling ability of global information. In the feedforward neural network, a more effective Bilinear layer is employed for processing to capture the higher-order relationship between input features, thereby enhancing the ability of the model to capture complex patterns. Experiments conducted on the RML2016.10A public dataset demonstrate that compared with the existing algorithms, the proposed algorithm not only exhibits significant advantages in terms of parameter efficiency but also achieves higher recognition accuracy under various signal-to-noise ratio (SNR) conditions. In particular, it performs relatively well in terms of accuracy, precision, recall, and F1-score, with clearer classification of higher-order modulations and notable overall accuracy improvement.

Impact of internal phase volume on the physical, morphological and mechanical characteristics of emulsion templated scaffolds

Objectives: The high porosity of tissue engineering scaffolds is advantageous as they provide a high degree of infiltration of nutrients, enable cell penetration, and support vascularisation. However, the mechanical strength is also critical for providing structural support to the defect site throughout the regeneration process. In this study, we aimed to establish a relationship between internal phase volume and emulsion-templated scaffolds' physical, morphological and mechanical characteristics. Methods: In this work, tetra methacrylate functionalised polycaprolactone (4PCLMA) polymers were synthesised via ring-opening polymerisation followed by methacrylation. 4PCLMA-based emulsion templated matrices with 60%, 75% and 82% internal phase volumes were fabricated (P60, P75, and P82). These scaffolds' densities, porosities, average pore and window sizes, degree of interconnectivity values, and mechanical properties were investigated. Results: Increasing internal phase volume reduced the density of the foams by almost two-fold. No direct correlation was observed between average pore size and internal phase volume. Both the average window sizes and the degree of interconnectivity values increase with increasing internal phase volume. Compression modulus values are calculated as 0.46±0.04 MPa, 0.23±0.02 MPa and 0.14±0.01 MPa for P60, P75, and P82, respectively. Increasing internal phase volume from 60% to 82% caused a more than 2-fold reduction in the stiffness of the emulsion-templated matrices. Conclusions: Accordingly, by reporting on this experimental framework, we established a relationship between internal phase volume and the physical, morphological and mechanical characteristics of 4PCMA-based scaffolds to precisely engineer these characteristics for specific tissue engineering applications.

Optimizing real-time phase detection in diverse rhythmic biological signals for phase-specific neuromodulation.

Closed-loop, phase-specific neurostimulation is a powerful method to modulate ongoing brain activity for clinical and research applications. Phase-specific stimulation relies on estimating the phase of an ongoing oscillation in real time and issuing a control command at a target phase. Phase detection algorithms based on Fast Fourier transform (FFT) are widely used due to their computational efficiency and robustness. However, it is unclear how algorithm performance depends on the spectral properties of the input signal and how algorithm parameters can be optimized. We used offline simulation to evaluate the performance of three algorithms (endpoint-corrected Hilbert Transform, Hilbert Transform and phase mapping) on three rhythmic biological signals with distinct spectral properties (rodent hippocampal theta potential, human EEG alpha and human essential tremor). First, we found that algorithm performance was more strongly influenced by signal amplitude and frequency variation compared with signal to noise ratio. Second, our simulations showed that the size of the data window for phase estimation was critical for the performance of FFT-based algorithms, where the optimal data window corresponds to the period of the oscillation. We validated this prediction with real time phase detection of hippocampal theta oscillations in freely behaving rats performing spatial navigation. Our findings define the relationship between signal properties and algorithm performance and provide a convenient method for optimizing FFT-based phase detection algorithms.

TPANet: A novel triple parallel attention network approach for remaining useful life prediction of lithium-ion batteries

Accurate remaining useful life (RUL) prediction for lithium-ion batteries (LIBs) is important for proper equipment operation. Among the numerous existing studies on battery research, data-driven approaches have gained significant attention because they obviate the need for complex chemical and physical modeling of battery processes. However, in most current data-driven methods, the sliding window size values are determined empirically, which affects both the time required for model prediction and the prediction accuracy. To address this issue, this paper proposes a two-stage prediction method for battery capacity aging trajectories. In the first stage, a false nearest neighbor (FNN) technique is utilized to infer the sliding window size of the battery, reducing the error associated with manually selecting the sliding window size. In the second stage, a novel triple parallel attention network (TPANet) based on convolutional neural networks (CNNs) and attention units is developed, which extracts the input battery capacity degradation features through a parallel mechanism. The introduction of attention units in each parallel branch allows the model to enhance the capture of capacity regeneration phenomena as well as long-term dependence on input data. The proposed approach is validated on two publicly available datasets as well as a battery developed by ourselves. Diverse simulation results demonstrate the ability of the proposed approach to accurately predict the RUL of LIBs in a short time with broad generalization capabilities.

Microporous membrane with ionized sub-nanochannels enabling highly selective monovalent and divalent anion separation

Membranes tailored for selective ion transport represent a promising avenue toward enhancing sustainability across various fields including water treatment, resource recovery, and energy conversion and storage. While nanochannels formed by polymers of intrinsic microporosity (PIM) offer a compelling solution with their uniform and durable nanometer-sized pores, their effectiveness is hindered by limited interactions between ions and nanochannel. Herein, we introduce the randomly twisted V-shaped structure of Tröger’s Base unit and quaternary ammonium groups to construct ionized sub-nanochannel with a window size of 5.89–6.54 Å between anion hydration and Stokes diameter, which enhanced the dehydrated monovalent ion transport. Combining the size sieving and electrostatic interaction effects, sub-nanochannel membranes achieved exceptional ion selectivity of 106 for Cl-/CO32- and 82 for Cl-/SO42-, significantly surpassing the state-of-the-art membranes. This work provides an efficient template for creating functionalized sub-nanometer channels in PIM membranes, and paves the way for the development of precise ion separation applications.

Digital bio-dosimeter for detection of deformed biological cells irradiated by ionizing radiations

In this paper, a digital bio-dosimeter (DBD) with a square window size of 7 × 7 pixels is proposed for the quantitative detection of the deformed biological cells irradiated by X-rays with different energies and γ-rays from 60Co. The basic idea of the DBD is to compute the energy of the bio-dosimeter window size by summing up the entire gray levels. Peripheral blood films (PBFs) from male albino rats irradiated by ionizing radiations are prepared and photographed using an Automatic Image Contour Analysis system with high magnification to produce a 2D image. The surf function in MATLAB is used to convert the 2D image to a 3D shaded surface to resample the date more finely. The proposed DBD shows that the computed energies on the boundaries of the normal RBCs are nearly the same, while the computed energies on the boundaries of the deformed cells are significantly larger.

Schizophrenia diagnosis based on diverse epoch size resting-state EEG using machine learning.

Schizophrenia is a severe mental disorder that impairs a person's mental, social, and emotional faculties gradually. Detection in the early stages with an accurate diagnosis is crucial to remedying the patients. This study proposed a new method to classify schizophrenia disease in the rest state based on neurologic signals achieved from the brain by electroencephalography (EEG). The datasets used consisted of 28 subjects, 14 for each group, which are schizophrenia and healthy control. The data was collected from the scalps with 19 EEG channels using a 250 Hz frequency. Due to the brain signal variation, we have decomposed the EEG signals into five sub-bands using a band-pass filter, ensuring the best signal clarity and eliminating artifacts. This work was performed with several scenarios: First, traditional techniques were applied. Secondly, augmented data (additive white Gaussian noise and stretched signals) were utilized. Additionally, we assessed Minimum Redundancy Maximum Relevance (MRMR) as the features reduction method. All these data scenarios are applied with three different window sizes (epochs): 1, 2, and 5 s, utilizing six algorithms to extract features: Fast Fourier Transform (FFT), Approximate Entropy (ApEn), Log Energy entropy (LogEn), Shannon Entropy (ShnEn), and kurtosis. The L2-normalization method was applied to the derived features, positively affecting the results. In terms of classification, we applied four algorithms: K-nearest neighbor (KNN), support vector machine (SVM), quadratic discriminant analysis (QDA), and ensemble classifier (EC). From all the scenarios, our evaluation showed that SVM had remarkable results in all evaluation metrics with LogEn features utilizing a 1-s window size, impacting the diagnosis of Schizophrenia disease. This indicates that an accurate diagnosis of schizophrenia can be achieved through the right features and classification model selection. Finally, we contrasted our results to recently published works using the same and a different dataset, where our method showed a notable improvement.

Modeling and Performance Analysis of LBT-Based RF-Powered NR-U Network for IoT.

Energy harvesting combined with spectrum sharing offers a promising solution to the growing demand for spectrum while keeping energy costs low. New Radio Unlicensed (NR-U) technology enables telecom operators to utilize unlicensed spectrum in addition to the licensed spectrum already in use. Along with this, the energy demands for the Internet of Things (IoT) can be met through energy harvesting. In this regard, the ubiquity and ease of implementation make the RF-powered NR-U network a sustainable solution for cellular IoT. Using a Markov chain, we model the NR-U network with nodes powered by the base station (BS). We derive closed-form expressions for the normalized saturated throughput of nodes and the BS, along with the mean packet delay at the node. Additionally, we compute the transmit outage probability of the node. These quality of service (QoS) parameters are analyzed for different values of congestion window size, TXOP parameter, maximum energy level, and energy threshold of the node. Additionally, the effect of network density on collision, transmission, and energy harvesting probabilities is observed. We validate our model through simulations.