Computational Complexity Research Articles

Gesture recognition from surface electromyography signals based on the SE-DenseNet network.

In recent years, significant progress has been made in the research of gesture recognition using surface electromyography (sEMG) signals based on machine learning and deep learning techniques. The main motivation for sEMG gesture recognition research is to provide more natural, convenient, and personalized human-computer interaction, which makes research in this field have considerable application prospects in rehabilitation technology. However, the existing gesture recognition algorithms still need to be further improved in terms of global feature capture, model computational complexity, and generalizability. This paper proposes a fusion model of Squeeze-and-Excitation Networks (SE) and DenseNet, inserting attention mechanism between DenseBlock and Transition to focus on the most important information, improving feature representation ability, and effectively solving the problem of gradient vanishing. This proposed method was tested on the electromyographic gesture datasets NinaPro DB2 and DB4, achieving accuracies of 85.93 and 82.39 % respectively. Through ablation experiments, it was found that the method based on DenseNet-101 as the backbone model produced the best results. Compared with existing models, this proposed method has better robustness and generalizability in gesture recognition, providing new ideas for the development of sEMG signal gesture recognition applications in the future.

Electrical impedance tomography meets reduced order modelling: a framework for faster and more reliable electrical conductivity estimations

Objective.Inclusion of individualised electrical conductivities of head tissues is crucial for the accuracy of electrical source imaging techniques based on electro/magnetoencephalography and the efficacy of transcranial electrical stimulation. Parametric electrical impedance tomography (pEIT) is a method to cheaply and non-invasively estimate them using electrode arrays on the scalp to apply currents and measure the resulting potential distribution. Conductivities are then estimated by iteratively fitting a forward model to the measurements, incurring a prohibitive computational cost that is generally lowered at the expense of accuracy. Reducing the computational cost associated with the forward solutions would improve the accessibility of this method and unlock new capabilities.Approach.We introduce reduced order modelling (ROM) to massively speed up the calculations of these solutions for arbitrary conductivity values.Main results.We demonstrate this new ROM-pEIT framework using a realistic head model with six tissue compartments, with minimal errors in both the approximated numerical solutions and conductivity estimations. We show that the computational complexity required to reach a multi-parameter estimation with a negligible relative error is reduced by more than an order of magnitude when using this framework. Furthermore, we illustrate the benefits of this new framework in a number of practical cases, including its application to real pEIT data from three subjects.Significance.Results suggest that this framework can transform the use of pEIT for seeking personalised head conductivities, making it a valuable tool for researchers and clinicians.

A novel arc detection and identification method in pantograph-catenary system based on deep learning

Arc detection is crucial for ensuring the safe operation of power systems, where timely and accurate detection of arcs can prevent potential hazards such as fires, equipment damage, or system failures. Traditional arc detection methods, while functional, often suffer from low detection accuracy and high computational complexity, especially in complex operational environments. This limitation is particularly problematic in real-time monitoring and the efficient operation of power systems. In order to solve these problems, this paper proposes a method of arc detection based on deep learning, called arc multi-scene detection (ArcMSD), which leverages deep learning techniques to address these challenges. The primary distinction of this method lies in its enhancement of the Inception V3 model. This paper has redesigned the original Inception module by incorporating a guided anchor mechanism, an attention mechanism, and upsampling techniques to optimize detection performance. The improved Inception V3 network uses an attention mechanism to allow the model to focus on arc features in complex backgrounds, which can also prevent the model from overfitting. It performs upsampling and fusion with low-level features in the model. The fused features have better arc discrimination capabilities than the original input features, which better improves the accuracy of the model. In order to adapt to arcs with large size differences and improve detection efficiency, the guided anchor is selected to adjust the anchor generation algorithm. In terms of dataset, continuous frame images are intercepted from the video of Integrated Supervision and Control System (ISCS), and image preprocessing operations are performed to improve the model’s detection accuracy of pantograph arcs. Experimental results show that the mean Average Precision (mAP) of the deep learning model proposed in this article is 95.4%, which is far better than other models, thus verify the method’s efficacy.

Novel techniques for efficient PAPR reduction in NOMA systems for future wireless networks

ABSTRACT Fifth-generation (5G) networks are designed to overcome critical challenges, including achieving high data rates, supporting massive device connectivity, and ensuring low latency. Non-Orthogonal Multiple Access (NOMA) significantly enhances spectral efficiency and network capacity within this framework. NOMA has a problem with a high Peak-to-Average Power Ratio (PAPR). This creates nonlinear distortion, reduces power amplifier efficiency, and increases out-of-band radiation, making it less useful in 5G and B5G networks. To address this, two PAPR reduction methods are proposed: Improved Salp Swarm Algorithm-based Partial Transmit Sequence (PTS-ISSA) and Iterative Sub-block Phase Rotation (ISPR). PTS-ISSA minimizes PAPR by phase rotation of sub-blocks, offering the best reduction but at higher computational complexity, making it ideal for systems with abundant resources. ISPR, on the other hand, employs iterative phase rotation without side information, balancing performance with lower complexity, making it suitable for resource-constrained applications. Evaluated under Rician fading and AWGN channels, both methods outperform traditional techniques in PAPR reduction and Bit Error Rate (BER) performance. While PTS-ISSA is optimal for high-performance systems, ISPR serves lightweight, real-time applications. These methodologies show potential for the next-generation wireless networks, supporting scenarios from IoT to ultra-reliable low-latency communications by optimizing the trade-off between performance and complexity.

Low complexity neural network equalizer for nonlinearity mitigation in digital subcarrier multiplexing systems.

This article conducts a comparative study of the complexity reduction of neural network (NN) models for nonlinearity compensation used in digital subcarrier multiplexing (DSCM)-based optical communication systems. We employ the NN model based on bi-directional long short-term memory (biLSTM) and 1D-convolutional NN (1D-CNN) layers. To reduce the computation complexity of the proposed solution, weight clustering is applied to the NN. We specifically compare the performance of our proposed NN-based equalizer with traditional methods such as chromatic dispersion compensation (CDC), digital back-propagation (DBP) and alternative NN, based on triplet coefficients in perturbation analysis, proposed in previous literature, evaluating both Q-factor performance and computational complexity. Our findings show that the NN-based equalizer offers competitive Q-factor improvements while significantly reducing computational complexity, particularly when weight clustering is used. We show that the complexity of the NN can be reduced by up to 91.1% compared to the NN based on the perturbation analysis proposed in previous literature and by 31.5% compared to the DBP with 1 step per span. The results underscore the potential of NN-based approaches to deliver high-performance nonlinear compensation with lower computational demands, positioning them as a promising solution for future optical nonlinear equalizers.

A comparative study of the performance of ten metaheuristic algorithms for parameter estimation of solar photovoltaic models

This study conducts a comparative analysis of the performance of ten novel and well-performing metaheuristic algorithms for parameter estimation of solar photovoltaic models. This optimization problem involves accurately identifying parameters that reflect the complex and nonlinear behaviours of photovoltaic cells affected by changing environmental conditions and material inconsistencies. This estimation is challenging due to computational complexity and the risk of optimization errors, which can hinder reliable performance predictions. The algorithms evaluated include the Crayfish Optimization Algorithm, the Golf Optimization Algorithm, the Coati Optimization Algorithm, the Crested Porcupine Optimizer, the Growth Optimizer, the Artificial Protozoa Optimizer, the Secretary Bird Optimization Algorithm, the Mother Optimization Algorithm, the Election Optimizer Algorithm, and the Technical and Vocational Education and Training-Based Optimizer. These algorithms are applied to solve four well-established photovoltaic models: the single-diode model, the double-diode model, the triple-diode model, and different photovoltaic module models. The study focuses on key performance metrics such as execution time, number of function evaluations, and solution optimality. The results reveal significant differences in the efficiency and accuracy of the algorithms, with some algorithms demonstrating superior performance in specific models. The Friedman test was utilized to rank the performance of the various algorithms, revealing the Growth Optimizer as the top performer across all the considered models. This optimizer achieved a root mean square error of 9.8602187789E−04 for the single-diode model, 9.8248487610E−04 for both the double-diode and triple-diode models and 1.2307306856E−02 for the photovoltaic module model. This consistent success indicates that the Growth Optimizer is a strong contender for future enhancements aimed at further boosting its efficiency and effectiveness. Its current performance suggests significant potential for improvement, making it a promising focus for ongoing development efforts. The findings contribute to the understanding of the applicability and performance of metaheuristic algorithms in renewable energy systems, providing valuable insights for optimizing photovoltaic models.

Identifying vital spreaders in large-scale networks based on neighbor multilayer contributions

IntroductionIdentifying influential spreaders in complex networks is crucial for understanding information propagation and disease immunity. The spreading ability of a node has been commonly assessed through its neighbor information. However, current methods do not provide specific explanations for the role of neighbors or distinguish their individual contributions to the spread of information.MethodsTo address these limitations, we propose an efficient ranking algorithm that strictly distinguishes the contribution of each neighbor in information spreading. This method combines the count of common neighbors with the K-shell value of each node to produce its ranking. By integrating these two factors, our approach aims to offer a more precise measure of a node's influence within a network.ResultsExtensive experiments were conducted using Kendall’s rank correlation, monotonicity tests, and the Susceptible-Infected-Recovered (SIR) epidemic model on real-world networks. These tests demonstrated the effectiveness of our proposed algorithm in identifying influential spreaders accurately.DiscussionFurthermore, computational complexity analysis indicates that our algorithm consumes less time compared to existing methods, suggesting it can be efficiently applied to large-scale networks.

Stabilizing Decomposition of Multiparameter Persistence Modules

Abstract While decomposition of one-parameter persistence modules behaves nicely, as demonstrated by the algebraic stability theorem, decomposition of multiparameter modules is known to be unstable in a certain precise sense. Until now, it has not been clear that there is any way to get around this and build a meaningful stability theory for multiparameter module decomposition. We introduce new tools, in particular $$\epsilon $$ ϵ -refinements and $$\epsilon $$ ϵ -erosion neighborhoods, to start building such a theory. We then define the $$\epsilon $$ ϵ -pruning of a module, which is a new invariant acting like a “refined barcode” that shows great promise to extract features from a module by approximately decomposing it. Our main theorem can be interpreted as a generalization of the algebraic stability theorem to multiparameter modules up to a factor of 2r, where r is the maximal pointwise dimension of one of the modules. Furthermore, we show that the factor 2r is close to optimal. Finally, we discuss the possibility of strengthening the stability theorem for modules that decompose into pointwise low-dimensional summands, and pose a conjecture phrased purely in terms of basic linear algebra and graph theory that seems to capture the difficulty of doing this. We also show that this conjecture is relevant for other areas of multipersistence, like the computational complexity of approximating the interleaving distance, and recent applications of relative homological algebra to multipersistence.

Human upper limb kinematics using a novel algorithm in post-stroke patients.

Assessing the kinematics of the upper limbs is crucial for rehabilitation treatment, especially for stroke survivors. Nowadays, researchers use computer vision-based algorithms for Human motion analysis. However, specific challenges include less accuracy, increased computational complexity and a limited number of anatomical key points. This study aims to develop a novel algorithm using the MediaPipe framework to estimate five specific upper limb movements in stroke survivors. A single mobile camera recorded the movements on their affected side in a study involving 10 hemiplegic patients. The algorithm was then utilized to calculate the angles associated with each movement, and its accuracy was validated against standard goniometer readings, showing a mean bias within an acceptable range. Additionally, a Bland-Altman analysis demonstrated a 95% limit of agreement between the algorithm's results and those of the Goniometer, indicating reliable performance. The MediaPipe framework provides several advantages over other methods like OpenPose and PoseNet, such as several anatomical key points, improved precision and reduced execution time. This algorithm facilitates efficient measurement of upper limb movement angles in stroke survivors and allows for straightforward tracking of mobility improvements. Such innovative technology is a valuable tool for healthcare professionals assessing upper limb kinematics in rehabilitation settings.

Efficient vehicular counting via privacy-aware aggregation network

Abstract Vehicle counting is crucial for effective road planning and traffic management. Despite significant advancements that have been achieved with the development of deep learning technology, current counting models rely on large-scale parameters and substantial computational resources, which limits their practical application. Additionally, these methods are typically trained on large centralized datasets, which may result in inefficiencies for resource-constrained devices. Furthermore, inadequate privacy protection poses potential risks of personal information leakage. To address these issues, we introduce a lightweight counting network, privacy-aware aggregation network (PANet) for real-world application in this paper. In PANet, a pyramid feature enhancement module is built to aggregate multi-scale information and enhance key representation, while also optimizing the channel-wise output of the model to reduce computational complexity. Furthermore, a federated learning framework is implemented to distribute the computational load and safeguard user privacy. Experimental results on a wide range of counting benchmarks demonstrate the superior efficiency and accuracy of PANet. The code is available at https://github.com/sdut-jacheng/PANet.

Binocular stereo vision-based relative positioning algorithm for drone swarm

To address the challenges of high computational complexity and poor real-time performance in binocular vision-based Unmanned Aerial Vehicle (UAV) formation flight, this paper introduces a UAV localization algorithm based on a lightweight object detection model. Firstly, we optimized the YOLOv5s model using lightweight design principles, resulting in Yolo-SGN. This model achieves a 65.5% reduction in parameter count, a 62.7% reduction in FLOPs, and a 1.8% increase in accuracy compared to the original detection model. Subsequently, Yolo-SGN is utilized to extract target regions from binocular images, and feature point matching is exclusively conducted within these regions to minimize unnecessary computations in non-target areas. Experimental results demonstrate that the combination of Yolo-SGN and the Oriented FAST and Rotated BRIEF (ORB) algorithm reduces feature point matching computations to only a quarter of those in the original ORB algorithm, significantly enhancing real-time performance. Finally, the extracted feature points from UAVs are input into a binocular vision localization model to compute their three-dimensional coordinates. The average of the three-dimensional coordinates of all feature points is used to determine the three-dimensional position of the target UAV. Experimental results confirm that the UAV binocular vision localization algorithm, based on a lightweight object detection model, exhibits exceptional performance in terms of precision and real-time capabilities.

A Spiral Coverage Path Planning Algorithm for Nonomnidirectional Robots

ABSTRACTThe limited steering capabilities of nonomnidirectional robots introduce significant complexity into complete coverage tasks, often leading to increased path overlap or incomplete coverage of certain areas. Although recent research has made progress in optimizing coverage path planning, redundant coverage or omissions are still prone to occur in the target area to be covered. To address these persistent challenges, we propose a novel spiral coverage method. This approach not only conforms to the kinematic constraints of nonomnidirectional robots but also enhances coverage efficiency by dividing the target area into center and boundary regions and devising tailored coverage strategies for each. This method effectively reduces path redundancy and improves overall area coverage. Furthermore, we introduce a comprehensive metric that combines total path length and area coverage ratio to evaluate coverage efficiency, overcoming the limitations and computational complexity associated with existing metrics. For scenarios where maximizing the area coverage ratio is critical, we have developed a high‐coverage‐rate turning strategy that ensures 100% coverage. Through simulation tests in six representative areas and actual experiments on airport runways, our method shows an improvement of 0.238%–14.538% in coverage efficiency compared with parallel coverage method and 60.548%–76.339% compared with deep reinforcement learning‐based method. Additionally, implementing high‐coverage‐rate turning strategies improves the area coverage ratio by 2.021%–6.732%. In field experiments, our method reduces execution time by 1.61% compared with parallel coverage method. These results show that our method has a significant effect in improving coverage efficiency and achieving complete coverage goals.

A Coarse- and Fine-Grained Co-Exploration Approach for Optimizing DNN Spatial Accelerators: Improving Speed and Performance

The rapid advancement of deep neural networks has significantly increased demands for computational complexity and data volume. This trend is especially evident with the emergence of large language models, which have rendered traditional architectures such as CPUs and GPGPUs insufficient in meeting performance and energy efficiency requirements. Spatial accelerators present a promising solution by optimizing on-chip compute, storage, and communication resources. In exploring spatial accelerator design spaces, analytical model-based simulators and cycle-accurate simulators are commonly employed, each offering distinct advantages: high computational speed and superior simulation accuracy, respectively. However, the limited accuracy of analytical models and the slow simulation speed of cycle-accurate simulators impede the achievement of globally optimal solutions during design space exploration. Therefore, effectively leveraging the strengths of both simulator types while mitigating their inherent trade-offs is a critical challenge in designing customized spatial accelerators. In this work, we introduce a novel co-exploration methodology that integrates both coarse-grained and fine-grained approaches to navigate design and mapping spaces effectively. We utilize the rapid simulation capabilities of analytical models to perform coarse-grained global exploration, quickly eliminating designs and mapping configurations with inferior performance. Building on the results of this initial exploration, we employ cycle-accurate simulators to conduct fine-grained local exploration within the identified promising regions of the design and mapping spaces. This dual-phase approach aims to identify optimal hardware designs and dataflow mapping strategies that enhance performance and energy efficiency. The experimental results demonstrate that, compared to state-of-the-art methods, our approach reduces the number of exploration points by up to 99%, while achieving a 17.9% reduction in latency, a 2.5% decrease in energy consumption, and a 30.3% improvement in throughput.

Towards an Efficient Remote Sensing Image Compression Network with Visual State Space Model

In the past few years, deep learning has achieved remarkable advancements in the area of image compression. Remote sensing image compression networks focus on enhancing the similarity between the input and reconstructed images, effectively reducing the storage and bandwidth requirements for high-resolution remote sensing images. As the network’s effective receptive field (ERF) expands, it can capture more feature information across the remote sensing images, thereby reducing spatial redundancy and improving compression efficiency. However, the majority of these learned image compression (LIC) techniques are primarily CNN-based and transformer-based, often failing to balance the global ERF and computational complexity optimally. To alleviate this issue, we propose a learned remote sensing image compression network with visual state space model named VMIC to achieve a better trade-off between computational complexity and performance. Specifically, instead of stacking small convolution kernels or heavy self-attention mechanisms, we employ a 2D-bidirectional selective scan mechanism. Every element within the feature map aggregates data from multiple spatial positions, establishing a globally effective receptive field with linear computational complexity. We extend it to an omni-selective scan for the global-spatial correlations within our Channel and Global Context Entropy Model (CGCM), enabling the integration of spatial and channel priors to minimize redundancy across slices. Experimental results demonstrate that the proposed method achieves superior trade-off between rate-distortion performance and complexity. Furthermore, in comparison to traditional codecs and learned image compression algorithms, our model achieves BD-rate reductions of −4.48%, −9.80% over the state-of-the-art VTM on the AID and NWPU VHR-10 datasets, respectively, as well as −6.73% and −7.93% on the panchromatic and multispectral images of the WorldView-3 remote sensing dataset.

DCFA-YOLO: A Dual-Channel Cross-Feature-Fusion Attention YOLO Network for Cherry Tomato Bunch Detection

To better utilize multimodal information for agriculture applications, this paper proposes a cherry tomato bunch detection network using dual-channel cross-feature fusion. It aims to improve detection performance by employing the complementary information of color and depth images. Using the existing YOLOv8_n as the baseline framework, it incorporates a dual-channel cross-fusion attention mechanism for multimodal feature extraction and fusion. In the backbone network, a ShuffleNetV2 unit is adopted to optimize the efficiency of initial feature extraction. During the feature fusion stage, two modules are introduced by using re-parameterization, dynamic weighting, and efficient concatenation to strengthen the representation of multimodal information. Meanwhile, the CBAM mechanism is integrated at different feature extraction stages, combined with the improved SPPF_CBAM module, to effectively enhance the focus and representation of critical features. Experimental results using a dataset obtained from a commercial greenhouse demonstrate that DCFA-YOLO excels in cherry tomato bunch detection, achieving an mAP50 of 96.5%, a significant improvement over the baseline model, while drastically reducing computational complexity. Furthermore, comparisons with other state-of-the-art YOLO and other object detection models validate its detection performance. This provides an efficient solution for multimodal fusion for real-time fruit detection in the context of robotic harvesting, running at 52fps on a regular computer.

A Novel Data Obfuscation Framework Integrating Probability Density and Information Entropy for Privacy Preservation

Data privacy protection is increasingly critical in fields like healthcare and finance, yet existing methods, such as Fully Homomorphic Encryption (FHE), differential privacy (DP), and federated learning (FL), face limitations like high computational complexity, noise interference, and communication overhead. This paper proposes a novel data obfuscation method based on probability density and information entropy, leveraging a probability density extraction module for global data distribution modeling and an information entropy fusion module for dynamically adjusting the obfuscation intensity. In medical image classification, the method achieved precision, recall, and accuracy of 0.93, 0.89, and 0.91, respectively, with a throughput of 57 FPS, significantly outperforming FHE (0.82, 23 FPS) and DP (0.84, 25 FPS). Similarly, in financial prediction tasks, it achieved precision, recall, and accuracy of 0.95, 0.91, and 0.93, with a throughput of 54 FPS, surpassing traditional approaches. These results highlight the method’s ability to balance privacy protection and task performance effectively, offering a robust solution for advancing privacy-preserving technologies.

Exploring Dominating Functions and Their Complexity in Subclasses of Weighted Chordal Graphs and Bipartite Graphs

Domination problems are fundamental problems in graph theory with diverse applications in optimization, network design, and computational complexity. This paper investigates {k}-domination, k-tuple domination, and their total domination variants in weighted strongly chordal graphs and chordal bipartite graphs. Specifically, the {k}-domination problem in weighted strongly chordal graphs and the total {k}-domination problem in weighted chordal bipartite graphs are shown to be solvable in O(n+m) time. For weighted proper interval graphs and convex bipartite graphs, we solve the k-tuple domination and total k-tuple domination problems in O(n2.371552log2(n)log(n/δ)), where δ is the desired accuracy. Furthermore, for weighted unit interval graphs, the k-tuple domination problem achieves a significant complexity improvement, reduced from O(nk+2) to O(n2.371552log2(n)log(n/δ)). These results are achieved through a combination of linear and integer programming techniques, complemented by totally balanced matrices, totally unimodular matrices, and graph-specific matrix representations such as neighborhood and closed neighborhood matrices.

MESSI: Task Mapping and Scheduling Strategy for FPGA-based Heterogeneous Real-Time Systems

Continuous demands for improved performance within constrained resource budgets are driving a move from homogeneous to heterogeneous processing platforms for the implementation of today’s Real-Time (RT) embedded systems. The applications executing on such systems are typically represented as a Precedence Task Graph (PTG), where a node represents a task or algorithm for one functionality and edges represent the complex interactions between multiple functionalities. Due to RT constraints, the task graph needs to be executed within a specified deadline. Although some existing studies have looked into solving this challenge, comprehensive studies that combine the theoretical features of RT task-graph mapping and scheduling with practical runtime architectural characteristics have mostly been ignored to date. Hence, in this paper, we consider the challenge of scheduling a RT application modelled as a single PTG, with the objective of minimizing the overall execution time under Hardware (HW) resource and deadline constraints for heterogeneous Central Processing Unit (CPU) + Field Programmable Gate Array (FPGA) architectures. First, we introduce an optimal solution using Integer Linear Programming (ILP). However, this ILP-based optimal solution suffers from computational complexity and does not scale well even for moderately large problem sizes. Hence, we additionally propose heuristic algorithms for task mapping and scheduling. The efficiency of the proposed scheme, named MESSI, has been evaluated through experiments using PTGs on a practical CPU+FPGA system regarding current technology restrictions. Our experiments demonstrate that performance gains of 55.6 % and area usage reductions of 46.3 % are possible compared to full Software SW and HW execution, respectively.

3D medical image segmentation using the serial–parallel convolutional neural network and transformer based on cross‐window self‐attention

AbstractConvolutional neural network (CNN) with the encoder–decoder structure is popular in medical image segmentation due to its excellent local feature extraction ability but it faces limitations in capturing the global feature. The transformer can extract the global information well but adapting it to small medical datasets is challenging and its computational complexity can be heavy. In this work, a serial and parallel network is proposed for the accurate 3D medical image segmentation by combining CNN and transformer and promoting feature interactions across various semantic levels. The core components of the proposed method include the cross window self‐attention based transformer (CWST) and multi‐scale local enhanced (MLE) modules. The CWST module enhances the global context understanding by partitioning 3D images into non‐overlapping windows and calculating sparse global attention between windows. The MLE module selectively fuses features by computing the voxel attention between different branch features, and uses convolution to strengthen the dense local information. The experiments on the prostate, atrium, and pancreas MR/CT image datasets consistently demonstrate the advantage of the proposed method over six popular segmentation models in both qualitative evaluation and quantitative indexes such as dice similarity coefficient, Intersection over Union, 95% Hausdorff distance and average symmetric surface distance.

XTNSR: Xception-based transformer network for single image super resolution

Single image super resolution has significantly advanced by utilizing transformers-based deep learning algorithms. However, challenges still need to be addressed in handling grid-like image patches with higher computational demands and addressing issues like over-smoothing in visual patches. This paper presents a Deep Learning model for single-image super-resolution. In this paper, we present the XTNSR model, a novel multi-path network architecture that combines Local feature window transformers (LWFT) with Xception blocks for single-image super-resolution. The model processes grid-like image patches effectively and reduces computational complexity by integrating a Patch Embedding layer. Whereas the Xception blocks use depth-wise separable convolutions for hierarchical feature extraction, the LWFT blocks capture long-range dependencies and fine-grained qualities. A multi-layer feature fusion block with skip connections, part of this hybrid architecture, guarantees efficient local and global feature fusion. The experimental results show better performance in Peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and visual quality than the state-of-the-art techniques. By optimizing parameters, the suggested architecture also lowers computational complexity. Overall, the architecture presents a promising approach for advancing image super-resolution capabilities.