High Complexity Research Articles

Mental Workload as a Predictor of ATCO’s Performance: Lessons Learnt from ATM Task-Related Experiments

Air Traffic Controllers’ (ATCos) mental workload is likely to remain the specific greatest functional limitation on the capacity of the Air Traffic Management (ATM) system. Developing computational models to monitor mental workload and task complexity is essential for enabling ATCOs and ATM systems to adapt to varying task demands. Most methodologies have computed task complexity based on basic parameters such as air-traffic density; however, literature research has shown that it also depends on many other factors. In this paper, we present a study in which we explored the possibility of predicting task complexity and performance through mental workload measurements of participants performing an ATM task in an air-traffic control simulator. Our findings suggest that mental workload measurements better predict poor performance and high task complexity peaks than other established factors. This underscores their potential for research into how different ATM factors affect task complexity. Understanding the role and the weight of these factors in the overall task complexity confronted by ATCos constitutes one of the biggest challenges currently faced by the ATM sphere and would significantly contribute to the safety of our sky.

YOLO-SAG: An improved wildlife object detection algorithm based on YOLOv8n

Wildlife conservation is crucial for maintaining biodiversity, ensuring ecosystem balance and stability, and fostering sustainable development. Currently, the use of infrared camera traps to monitor and capture photos of wildlife is a vital methodology in protecting and researching wildlife, and automatic detection and identification of animals within captured photographs are paramount. However, factors such as the complexity of the field environment and the varying sizes of animal targets lead to low detection accuracy, while high-precision detection models are hindered by high computational complexity and sluggish training speeds. This paper proposes a wildlife target detection algorithm based on improved YOLOv8n - YOLO-SAG, which aims to balance accuracy and speed. Training stability is enhanced by introducing the Softplus activation function, which increases detection accuracy; incorporating the AIFI enhances intra-scale feature interaction, reducing missed and false detections. Integrating the GSConv and VoV-GSCSP module lightens neck convolutions, reducing computational redundancy and balancing the computational and parametric quantities brought by the AIFI. Experimental results on a self-made wildlife dataset indicate that the YOLO-SAG achieves 94.9%, 90.9%, 96.8%, and 79.9% in Precision, Recall, mAP@0.5, and mAP@0.5–0.95, respectively, which are 3.4%, 3.3%, 3.2%, and 4.9% higher than the original YOLOv8n. Inference and post-processing times reach 1.2 ms and 0.5 ms, a speedup of 25% and 54.5%, respectively, and the computation volume is only 7.2 GFLOPs, an 11.1% decrease.

A low-complexity ADMM-based secure beamforming for IRS-assisted uplink cognitive radio networks with NOMA

Intelligent reflecting surface (IRS) and non-orthogonal multiple access (NOMA) are both powerful means to improve the spectrum efficiency and can assist the cognitive radio network (CRN) in achieving more effective spectrum allocation. However, spectrum sharing in CRNs and the utilization of the scarce bandwidth in NOMA make it vulnerable to security threats. The secure beamforming design in IRS-assisted CRNs with NOMA is complicated and non-convex. Traditional semidefinite relaxation along with Gaussian randomization (SDR-GR) algorithm can solve the non-convex problem, however a high computational complexity. This paper proposes an alternating direction method of multipliers (ADMM)-based sum secrecy rate maximization (A-SSRM) scheme to achieve secure transmission with low computation complexity in IRS-assisted uplink CRNs with NOMA. Specifically, we derive an appropriate transformation of the original problem based on the ADMM framework and decompose it into multiple sub-problems, which are carefully designed to have closed-form solutions. Furthermore, an eavesdropper zero forcing (EZF)-based sub-optimal scheme is presented. Simulation results reveal that our proposed A-SSRM scheme outperforms other benchmarks in terms of sum secrecy rate and reduces the average running time by 6 times compared with the traditional SDR-GR algorithm. This is attributed to our effective decoupling and decomposing of the original problem, which enables each sub-problem to obtain a simple closed-form solution through derivation.

Graduated Autonomy of Laparoscopic Liver Resection Based on Liver Resection Complexity: A Western and Eastern Bi-Institution Study for Learning Curve.

Laparoscopic liver resection (LLR) requires a high degree of expertise in both hepatobiliary and minimally invasive surgery. Our group previously reported a 3-level LLR complexity classification based on intra-postoperative outcomes: grade I (low), grade II (intermediate), and grade III (high). We evaluated the learning curve effect in each complexity grade to assess the experience needed for a surgeon to safely progress through the grades. Patients who underwent LLR during 1994-2020 at the Institut Mutualiste Montsouris (IMM) and The University of Tokyo (UT) during 2008-2023 were included in the study. The learning curve for operating time was evaluated using the standardized (CUSUM) analysis for each complexity grade. A total of 503 patients (grade I, 198; grade II, 87; grade III, 218) at the IMM and 221 patients (grade I, 135; grade II, 57; grade III, 29) at the UT met the inclusion criteria. The CUSUM analysis showed that the deviation of operating time was found up to 40 cases for grade I resections, 30 cases for grade II resections, and 50 cases for grade III resections. By dividing cohorts based on these numbers for each group and each institution and labeling these cases as the pre-learning groups and the remaining as the post-learning group, surgical outcomes and postoperative complications were generally improved in the post-learning groups in both institutions. A gradual progression in LLR per complexity grade as follow: 40 cases of low grade I procedures before starting intermediate complexity grade II procedures, and 30 cases of intermediate complexity grade II procedures before starting high complexity grade III procedures may ensure a safe implementation of high complexity LLR procedures.

Random memristor-based dynamic graph CNN for efficient point cloud learning at the edge

The broad integration of 3D sensors into devices like smartphones and AR/VR headsets has led to a surge in 3D data, with point clouds becoming a mainstream representation method. Efficient real-time learning of point cloud data on edge devices is crucial for applications such as autonomous vehicles and embodied AI. Traditional machine learning models on digital processors face limitations, with software challenges like high training complexity, and hardware challenges such as large time and energy overheads due to von Neumann bottleneck. To address this, we propose a software-hardware co-designed random memristor-based dynamic graph CNN (RDGCNN). Software-wise, we transform point cloud into graph, and propose random EdgeConv for efficient hierarchical and topological features extraction. Hardware-wise, leveraging memristor’s intrinsic stochasticity and in-memory computing capability, we achieve significant reductions in training complexity and energy consumption. RDGCNN demonstrates high accuracy and efficiency across various point cloud tasks, paving the way for future edge 3D vision.

Lightweight multiobject ship tracking algorithm based on trajectory association and improved YOLOv7tiny

In response to various challenges, such as high computational complexity, large parameter counts, and frequent vessel ID switching in ship tracking models, we propose a trajectory association and improved YOLOv7tiny-based vessel tracking algorithm, named the vessel tracking model with YOLOv7tiny-Vessel and StrongSORT-Vessel (VTM-YS). The algorithm consists of two stages: vessel detection and vessel tracking. In the vessel detection stage, we propose a lightweight Fast-PP-LCNet backbone network for efficient feature extraction of ships. Then, a novel Slim Bidirectional Pyramid Aggregation Network (Slim-BiPANet) is constructed to effectively fuse multiscale vessel features. Furthermore, introducing a coordinate attention (CA) module enhances the focus on vessel detection regions. In the vessel tracking stage, we introduce the OSNet reidentification network and the NSA Kalman filter algorithm into the StrongSORT model to reduce the frequency of tracked vessel ID switches. The results indicate that, in terms of vessel detection, the improved YOLOv7tiny achieves a 1.5% increase in mAP50 and a 13.9% reduction in parameters compared to the baseline YOLOv7tiny. For vessel tracking, the StrongSORT model achieves a 3.5% improvement in HOTA, a 7.0% improvement in MOTA, and a 73% reduction in ID switches, meeting the requirements for lightweight design and tracking accuracy.

Image encryption algorithm based on DNA mutation and a novel four-dimensional hyperchaos

Abstract Aiming at the problem that insufficient complexity of ordinary multi-dimensional chaotic systems and the cumbersome design of encryption algorithms without excellent encryption effects. This paper constructs a four-dimensional hyperchaotic system with high Lyapunov exponent and complex dynamic behavior. We designed an encryption algorithm based on point mutation, mutation diffusion, and folding mutation in DNA mutations. During the encryption process, we perform point mutation transformation on the entire base sequence, then spread the mutations one by one starting from the second base of the sequence, and finally flip every four base sequences according to folding mutations. The images encrypted by this algorithm have a uniform grayscale histogram, high information entropy, and high key sensitivity. It can resist exhaustive attacks, noise attacks, cropping attacks, and differential attacks, and have a fast encryption speed.

High-coverage characterization and discovery of molecular markers for quality control of natural fragrant plant extracts using UPLC-HRMS-based untargeted metabolomics.

The chemical components of natural fragrant plant extracts are of high complexity, and the strategies for quality control (QC) and further discovery of fragrance mechanisms still need to be further investigated. This study integrated the strategies and methods of untargeted metabolomics and chemometrics and statistical modeling to attain the goal. The techniques of reversed-phase and HILIC analysis of ultra-performance liquid chromatography-high-resolution mass spectrometry (UPLC-HRMS) were simultaneously used to collect data in both positive and negative ion modes. The pattern analysis of fingerprints and discovery of characteristic molecular markers for QC analysis were comprehensively employed to reach in-depth analysis of the quality variation and discovery of differential molecules among natural fragrant plant extracts. The former uses fingerprint technique to analyze their overall similarities and differences, and the latter comprehensively discovers molecular substances characterizing the chemical characteristics of fragrant extracts with the help of metabolomics and univariate and multivariate methods. The findings are expected to be used as the molecular markers in product manufacturing, sales, and consumption to achieve accurate quality control and recognition of targeted molecules for potential quality monitoring using spectroscopy techniques. In this work, 27 natural fragrant extracts were applied as examples, and their chemical components were comprehensively analyzed with discovery of markers for quality control. After data integration, 1178 molecules were annotated, and 267 differential metabolite molecules with the values of variable importance in the projection (VIP) larger than 1.0 were found. The results show that the method proposed in this work is of great significance for high-coverage analysis, QC marker discovery, and aroma mechanism elucidation, which has potential applications in the areas of food, cosmetics, pharmaceuticals, tobacco, and others.

Mortality in Patients with Intertrochanteric Fracture Operated Between 2019-2021 in a High Complexity Institution

Mortality in Patients with Intertrochanteric Fracture Operated Between 2019-2021 in a High Complexity Institution

A Proof-of-Concept FPGA-Based Clock Signal Phase Alignment System

Phase alignment of periodic events between multiple systems is required in multiple fields and applications. Most of the existing solutions focus on either low frequency and relatively low accuracy or high complexity, high accuracy and precision. In contrast, this work aimed to develop an intermediate solution, supporting high frequencies and relatively high accuracy and precision, with relatively low complexity. A hypothetical concept and mathematical model is presented with a hardware test implementation based entirely on FPGA resources. Deliberate resource selection and utilization enables a significant simplification of calculations and, as a result, a reduction in logic resource utilization. The proposed concept was implemented and verified using the AMD/Xilinx Artix 7 35T FPGA platform.

The association between medication use and health-related quality of life in multimorbid older patients with polypharmacy.

To explore the association between medication use-related factors and health-related quality of life (HRQoL) in older hospitalised multimorbid patients with polypharmacy. This cross-sectional study used the intervention arm data of the OPERAM trial (hospitalised patients ≥ 70years with polypharmacy). HRQoL was assessed using the visual analogue scale (EQ-VAS) and the EQ-5D index score of the EuroQol questionnaire (EQ-5D-5L). Lower or higher EQ-VAS/EQ-5D was based on the median of the study population. Medication use-related factors included hyperpolypharmacy (≥ 10 medications), anticholinergic and sedative burden, appropriateness of medication (STOPP/START criteria), high-risk medication for hospital (re)admission, medication complexity and adherence. Multivariable logistic regression analysis was used to assess the association between medication use-related factors and HRQoL. A total of 955 patients were included (mean age 79years, 46% female, median EQ-VAS of 60, median EQ-5D of 0.60). Opioids use was associated with lower EQ-5D and EQ-VAS (aOR EQ-5D: 2.10; 95% CI 1.34-3.32, EQ-VAS: 1.59; 1.11-2.30). Hyperpolypharmacy (aOR 1.37; 1.05-1.80), antibiotics (aOR 1.64; 1.01-2.68) and high medication complexity (aOR 1.53; 1.10-2.15) were associated with lower EQ-VAS. A high anticholinergic and sedative burden (aOR 1.73; 1.11-2.69), presence of multiple prescribing omissions (aOR 1.94; 1.19-3.17) and benzodiazepine use (aOR 2.01; 1.22-3.35) were associated with lower EQ-5D. Especially in hyperpolypharmacy patients, high anticholinergic and sedative burden and medication complexity were associated with a lower HRQoL. Several medication use-related factors are significantly associated with a lower HRQoL in hospitalised older patients. Medication complexity is a novel factor, which should be considered when evaluating medication use of older patients with hyperpolypharmacy.

New mechanistic insights into CO2 electroreduction to methane on penta-octa-graphene catalyst

Electrocatalytic reduction of CO2 to produce high value-added products is an important way to achieve sustainable production of green energy, but it is limited by high complexity of the catalyst structure and reaction process, making the rational design of the catalyst and the targeted generation of specific products a huge challenge. Herein, based on the penta-octa-graphene (POG) with sp2 and sp3 hybrid as the prototype, we theoretically guide the design of CO2 electrocatalytic reduction (CO2RR) catalyst over pure POG and TM doping POG (TM@POG). Our mechanistic study highlights two key factors in new pathway for CH4 production and catalytic performance, i.e., TM provides electron transfer for the key intermediate *CHO, promoting the activation of CO2, and the type of protonation for CO2RR is determined by the charge accumulation on the O atom of the intermediates. This reduces the complexity of the reaction and provides a predictive effect on the formation of intermediates. More importantly, the proper values of the Gibbs free energy indicate that TM@POG is an ideal catalyst candidate, especially the Fe@POG, which is superior to Cu (211) material. On this basis, three simple descriptors (Bader charge, d-band center εd and descriptor φ) based on the inherent properties of atoms were constructed to reveal the underlying causes of the improvement of catalytic activity. Our results provide a new understanding of the ’structure-performance’ relationship of carbon-based electrocatalysts, thus providing a useful strategy for the design of catalysts for CO2RR.

Physician Modified Endograft for Ruptured Dissecting Aortic Arch Aneurysm.

Endovascular repair of thoracic aortic aneurysms (TAA) in elective settings has demonstrated successful clinical outcomes. However, life-threatening conditions such as rupture are more often managed with open surgical repair due to the high complexity of arch endovascular repair, lack of available off-the-shelf devices, and limited long-term data. A 49-year-old female with a recent history of prior ascending aortic repair for Type A10 aortic dissection presented with chest pain and dyspnea. Chest computed tomography angiogram (CTA) revealed acute bilateral pulmonary emboli and a 6.2cm post dissection aneurysm of the posterior aortic arch with the dissection extending to the right iliac artery. She was treated with thrombolysis and subsequently became hemodynamically unstable. Repeat CTA revealed a massive left hemithorax with concern for aortic arch rupture. Given significant cardiorespiratory compromise and recent open repair, she was considered unfit for redo open repair. Thoracic endovascular aortic repair (TEVAR) with a physician-modified endograft (PMEG) was planned. An Alpha Zenith endograft was modified adding an internal branch for the innominate artery and a fenestration for the left common carotid artery. The left subclavian artery was occluded with a microvascular plug and coil embolization up to the level of the vertebral artery. TEVAR PMEG extension to the celiac artery was performed followed by deployment of a Zenith dissection stent to the aortic bifurcation. Completion angiogram demonstrated successful aneurysm exclusion and patency of target vessels. Endovascular treatment of ruptured TAA with PMEGs is feasible. This approach may be an alternative for unfit patients for open repair in emergent settings.

Spatial-temporal dynamic evolution of lewy body dementia by metabolic PET imaging.

Lewy body dementia (LBD) is a neurodegenerative disease with high heterogeneity and complex pathogenesis. Our study aimed to use disease progression modeling to uncover spatial-temporal dynamic evolution of LBD in vivo, and to explore differential profiles of clinical features, glucose metabolism, and dopaminergic function among different evolution-related subtypes. A total of 123 participants (31 healthy controls and 92 LBD patients) who underwent 18F-FDG PET scans were retrospectively enrolled. 18F-FDG PET-based Subtype and Stage Inference (SuStaIn) model was established to illustrate spatial-temporal evolutionary patterns and categorize relevant subtypes. Then subtypes and stages were further related to clinical features, glucose metabolism, and dopaminergic function of LBD patients. This 18F-FDG PET imaging-based approach illustrated two distinct patterns of neurodegenerative evolution originating from the neocortex and basal ganglia in LBD and defined them as subtype 1 and subtype 2, respectively. There were obvious differences between subtypes. Compared with subtype 1, subtype 2 exhibited a greater proportion of male patients (P = 0.045) and positive symptoms such as visual hallucinations (P = 0.033) and fluctuating cognitions (P = 0.033). Cognitive impairment, metabolic abnormalities, dopaminergic dysfunction and progression were all more severe in subtype 2 (all P < 0.05). In addition, a strong association was observed between SuStaIn subtypes and two clinical phenotypes (Parkinson's disease dementia and dementia with Lewy bodies) (P = 0.005). Our findings based on 18F-FDG PET and data-driven model illustrated spatial-temporal dynamic evolution of LBD and categorized novel subtypes with different evolutionary patterns, clinical and imaging features in vivo. The evolution-related subtypes are associated with LBD clinical phenotypes, which supports the perspective of existence of distinct entities in LBD spectrum.

Enhancing liver tumor segmentation with UNet-ResNet: Leveraging ResNet’s power

BACKGROUND: Liver cancer poses a significant health challenge due to its high incidence rates and complexities in detection and treatment. Accurate segmentation of liver tumors using medical imaging plays a crucial role in early diagnosis and treatment planning. OBJECTIVE: This study proposes a novel approach combining U-Net and ResNet architectures with the Adam optimizer and sigmoid activation function. The method leverages ResNet’s deep residual learning to address training issues in deep neural networks. At the same time, U-Net’s structure facilitates capturing local and global contextual information essential for precise tumor characterization. The model aims to enhance segmentation accuracy by effectively capturing intricate tumor features and contextual details by integrating these architectures. The Adam optimizer expedites model convergence by dynamically adjusting the learning rate based on gradient statistics during training. METHODS: To validate the effectiveness of the proposed approach, segmentation experiments are conducted on a diverse dataset comprising 130 CT scans of liver cancers. Furthermore, a state-of-the-art fusion strategy is introduced, combining the robust feature learning capabilities of the UNet-ResNet classifier with Snake-based Level Set Segmentation. RESULTS: Experimental results demonstrate impressive performance metrics, including an accuracy of 0.98 and a minimal loss of 0.10, underscoring the efficacy of the proposed methodology in liver cancer segmentation. CONCLUSION: This fusion approach effectively delineates complex and diffuse tumor shapes, significantly reducing errors.

SDCB-YOLO: A High-Precision Model for Detecting Safety Helmets and Reflective Clothing in Complex Environments

The correct wearing of safety helmets and reflective vests is of great significance in construction sites, offices, and civil engineering sites. Aiming to address the issues of low detection accuracy and high algorithm complexity caused by complex background environments in the small target detection of safety helmets and reflective clothing using existing algorithms, an improved algorithm based on YOLOv8n is proposed. Firstly, the SE module is utilized to reduce interference in complex environments. Next, the IOU function is modified to speed up calculations. Then, a lightweight universal upsampling operator (CARAFE) is employed to obtain a larger receptive field. Finally, the Bidirectional Feature Pyramid Network is used to replace the Concat module of the original head layer. Based on these four modifications made to the model, this article names the new model SDCB-YOLO, derived from the initial letters of the four respective modules. The experimental results show that the mAP of the SDCB-YOLO model on the test set reached 97.1%, which is 4.6% higher than YOLOv5s and 3.5% higher than YOLOv8n. Additionally, the model boasts a parameter count of 3,094,304, a computational load of 8.4 GFLOPs, and a model size of 6.13 MB. Compared to YOLOv5s, with a parameter count of 7,030,417, a computational cost of 16.0 GFLOPs, and a model size of 13.79 MB, the SDCB-YOLO model is significantly smaller. When compared to YOLOv8n, with a parameter count of 3,011,628, a computational complexity of 8.2 GFLOPs, and a model size of 6.11 MB, the SDCB-YOLO model’s parameters and model size are only slightly increased, while maintaining a comparable computational load. Therefore, the improved detection algorithm presented in this article not only ensures the lightweight nature of the model but also significantly enhances its detection accuracy.

Soft Fibrous Syringe Architecture for Electricity-Free and Motorless Control of Flexible Robotic Systems.

Flexible robotic systems (FRSs) and wearable user interfaces (WUIs) have been widely used in medical fields, offering lower infection risk and shorter recovery, and supporting amiable human-machine interactions (HMIs). Recently, soft electric, thermal, magnetic, and fluidic actuators with enhanced safety and compliance have innovatively boosted the use of FRSs and WUIs across many sectors. Among them, soft hydraulic actuators offer great speed, low noise, and high force density. However, they currently require bulky electric motors/pumps, pistons, valves, rigid accessories, and complex controllers, which inherently result in high cost, low adaptation, and complex setups. This paper introduces a novel soft fibrous syringe architecture (SFSA) consisting of two or more hydraulically connected soft artificial muscles that enable electricity-free actuation, motorless control, and built-in sensing ability for use in FRSs and WUIs. Its capabilities are experimentally demonstrated with various robotic applications including teleoperated flexible catheters, cable-driven continuum robotic arms, and WUIs. In addition, its sensing abilities to detect passive and active touch, surface texture, and object stiffness are also proven. These excellent results demonstrate a high feasibility of using a current-free and motor-less control approach for the FRSs and WUIs, enabling new methods of sensing and actuation across the robotic field.

Quantitative evaluation and genome-wide association studies of chrysanthemum flower color

Chrysanthemum, one of the economically most important and highly favored ornamental crops, exhibits a wide range of flower colors that are pivotal in determining its aesthetic value and market demand. However, probably due to the high genetic complexity of flower color traits, the progress in the dissection of genetic architecture and identification of quantitative trait nucleotides (QTNs) or candidate genes moves rather slowly. In this study, we systematically assessed the genetic variation of flower color phenotypes by integrating multiple color systems (RHSCC, CIELAB, and RGB) and four principal floral pigments (anthocyanins, chlorophylls, carotenoids, and flavonoids) in a diverse collection of 141 chrysanthemums. The results showed that the flower color traits exhibited a wide range of variations, with coefficients of variation (CV) varying from 26.00 % to 203.52 %. Among the 66 pairs of flower color traits, 48 exhibited highly significant correlations at p < 0.01 or p < 0.05 levels. Moreover, eight exponential regression models were established for predicting anthocyanin and carotenoid contents. Based on the 370,562 high-quality SNPs, genome-wide association studies (GWAS) revealed 36 QTNs significantly associated with flower color phenotype traits and 109 QTNs for pigment content traits. By integrating the functional annotation and expression profiles, a total of 809 candidate genes surrounding the QTNs were initially filtered out. The candidate genes encompassed multiple transcription factor families related to flavonoid metabolism pathways, including MYB, bHLH, WD40, NAC, and AP2/ERF transcription factors, as well as cytochrome P450-type enzyme genes. The current findings contribute to the accurate assessment of flower colors and provide valuable resources for the molecular breeding and functional analysis of flower color in chrysanthemums.

Impact of freeze-thaw cycles on the physicochemical properties and structure-function relationship of potato starch with varying granule sizes in frozen dough

Starch, as a critical component of dough, significantly influences quality preservation during the freezing process. In particular, the fine structure of potato (B-type) starch in frozen processing is a subject of considerable interest. This study aims to investigate the intrinsic differences of B-type starch and the impact of freeze-thaw (F/T) treatment on its molecular structure and physicochemical properties. Chain length distribution and X-ray photoelectron spectroscopy were utilized to examine the structural characteristics of natural potato starch with different granule sizes. Furthermore, the fine structure, thermal properties, and rheological properties of the isolated starches after F/T treatment were analyzed. The results indicate that potato starch with smaller particle sizes exhibits higher surface CC and PO content along with a higher proportion of very short chains (DP < 6, 8.17 %) and long B chains (DP > 25, 20.68 %). The study found that after F/T treatment, the surface of small-sized starch granules was initially damaged, exhibiting threads on the surface centered on the umbilical point. Following F/T treatment, both the crystallinity (very large (VL): 24.52–18.36 %; small (S): 17.03–16.69 %) and short-range order (VL: 2.97–2.61; S: 2.71–2.35) of starch particle size decreased. Both the amylose content (20.88–14.57 %) and ΔH (10.15–8.62 J/g) of isolated starch after freeze-thaw-treated dough exhibited a decrease to varying degrees. With the exception of the fifth cycle, small-size starch particles exhibited relatively higher G' and G" values and showed significant changes as a result of F/T treatment, demonstrating high hardness and complex viscosity. Clarifying the physicochemical properties of potato starches with different granule sizes is expected to expand their applications in frozen dough.

Intra-Pulse Modulation Recognition of Radar Signals Based on Efficient Cross-Scale Aware Network.

Radar signal intra-pulse modulation recognition can be addressed with convolutional neural networks (CNNs) and time-frequency images (TFIs). However, current CNNs have high computational complexity and do not perform well in low-signal-to-noise ratio (SNR) scenarios. In this paper, we propose a lightweight CNN known as the cross-scale aware network (CSANet) to recognize intra-pulse modulation based on three types of TFIs. The cross-scale aware (CSA) module, designed as a residual and parallel architecture, comprises a depthwise dilated convolution group (DDConv Group), a cross-channel interaction (CCI) mechanism, and spatial information focus (SIF). DDConv Group produces multiple-scale features with a dynamic receptive field, CCI fuses the features and mitigates noise in multiple channels, and SIF is aware of the cross-scale details of TFI structures. Furthermore, we develop a novel time-frequency fusion (TFF) feature based on three types of TFIs by employing image preprocessing techniques, i.e., adaptive binarization, morphological processing, and feature fusion. Experiments demonstrate that CSANet achieves higher accuracy with our TFF compared to other TFIs. Meanwhile, CSANet outperforms cutting-edge networks across twelve radar signal datasets, providing an efficient solution for high-precision recognition in low-SNR scenarios.