Search

In this work, we introduce a differentiable implementation of the local natural orbital coupled cluster (LNO-CC) method within the automatic differentiation framework of the PySCFAD package. The implementation is comprehensively tuned for enhanced performance, which enables the calculation of first-order static response properties on medium-sized molecular systems using coupled cluster theory with single, double, and perturbative triple excitations [CCSD(T)]. We evaluate the accuracy of our method by benchmarking it against the canonical CCSD(T) reference for nuclear gradients, dipole moments, and geometry optimizations. In addition, we demonstrate the possibility of property calculations for chemically interesting systems through the computation of bond orders and Mössbauer spectroscopy parameters for a [NiFe]-hydrogenase active site model, along with the simulation of infrared spectra via ab initio LNO-CC molecular dynamics for a protonated water hexamer. [ABSTRACT FROM AUTHOR]

This paper introduces a method to quickly and accurately identify the sound quality of automatic transmission in a vehicle by testing the objective vibration parameters on automatic transmission EOL (end of line) test bench. In this study, 50 automatic transmissions serve as the research object, and the vibration parameters of each gear of the transmission are measured on the EOL bench. Then, the transmission is mounted on a vehicle, the vibration parameters and acoustic parameters of corresponding working conditions are measured, and the subjective evaluation of the noise quality of the vehicle is carried out by using a grade scoring method. After this data is collected, the specific vibration parameters and acoustic parameters that are the dominant factors of sound quality are determined by correlation analysis. Based on the multiple linear regression method, the following mathematical models are formulated: the vibration parameters and the subjective evaluation of a vehicle, and the acoustic parameters and the subjective evaluation of a vehicle. Among these models, the mathematical models of the vibration parameters and the subjective evaluation of a vehicle were verified to be more accurate. The consistency between the vibration value in a vehicle and the vibration value on EOL test bench is analyzed and the EOL vibration value is determined to be the most objective evaluation of the data. After batch testing of 100 sets, the effectiveness of this model is within 95%. Finally, the value of critical amplitude of the condition of each gear of transmission is analyzed to form the judgment standard of transmission EOL vibration to ensure better sound quality for the vehicle. [ABSTRACT FROM AUTHOR]

The widespread application of virtual reality has made automatic target recognition in environmental space particularly important. However, traditional image recognition methods faced challenges due to the complex and diverse environmental characteristics in virtual reality scenes. It is easily influenced by occlusion, noise, and other factors. To deal with the issue, this paper utilizes the combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs), and proposes a collaborative neural network-based automatic object recognition method for environmental space under virtual reality scenes. Specifically, CNN is mainly responsible for feature extraction of environmental scene images, while RNN is used to model the spatial relationships of image sequences. Then, a data augmentation method is introduced in virtual reality scenarios, which can generate more training samples on a limited dataset, thereby improving the model’s generalization ability. Meanwhile, in order to address issues such as occlusion and noise, a multi-scale feature fusion strategy is further developed to improve the robustness of target recognition by multiscale feature fusion. Through experimental verification using actual data collected in virtual reality scenes, the results show that the proposed method has achieved excellent performance and strong adaptability compared with traditional methods. [ABSTRACT FROM AUTHOR]

This paper introduces the Walrus Optimization Algorithm (WaOA) to address load frequency control and automatic voltage regulation in a two-area interconnected power systems. The load frequency control and automatic voltage regulation are critical for maintaining power quality by ensuring stable frequency and voltage levels. The parameters of fractional order Proportional-Integral-Derivative (FO-PID) controller are optimized using WaOA, inspired by the social and foraging behaviors of walruses, which inhabit the arctic and sub-arctic regions. The proposed method demonstrates faster convergence in frequency and voltage regulation and improved tie-line power stabilization compared to recent optimization algorithms such as salp swarm, whale optimization, crayfish optimization, secretary bird optimization, hippopotamus optimization, brown bear optimization, teaching learning optimization, artificial gorilla troop optimization, and wild horse optimization. MATLAB simulations show that the WaOA-tuned FO-PID controller improves frequency regulation by approximately 25%, and exhibits a considerable faster settling time. Bode plot analyses confirm the stability with gain margins of 5.83 dB and 9.61 dB, and phase margins of 10.8 degrees and 28.6 degrees for the two areas respectively. The system modeling and validation in MATLAB showcases the superior performance and reliability of the WaOA-tuned FO-PID controller in enhancing power system stability and quality under step, random step load disturbance, with nonlinearities like GDC and GDB, and system parameter variations. [ABSTRACT FROM AUTHOR]

AbstractObjectiveDesignStudy sampleResultsConclusionsExploring applications of automatic speech recognition and text-to-speech technologies in hearing assessment and evaluations of hearing aids.Review protocol was registered at the INPLASY database and was performed following the PRISMA scoping review guidelines. A search in ten databases was conducted in January 2023 and updated in June 2024.Studies that used automatic speech recognition or text-to-speech to assess measures of hearing ability (e.g. speech reception threshold), or to configure hearing aids were retrieved. Of the 2942 records found, 28 met the inclusion criteria.The results indicated that text-to-speech could effectively replace recorded stimuli in speech intelligibility tests, requiring less effort for experimenters, without negatively impacting outcomes (n = 5). Automatic speech recognition captured verbal responses accurately, allowing for reliable speech reception threshold measurements without human supervision (n = 7). Moreover, automatic speech recognition was employed to simulate participants’ hearing, with high correlations between simulated and empirical data (n = 14). Finally, automatic speech recognition was used to optimise hearing aid configurations, leading to higher speech intelligibility for wearers compared to the original configuration (n = 3).There is the potential for automatic speech recognition and text-to-speech systems to enhance accessibility of, and efficiency in, hearing assessments, offering unsupervised testing options, and facilitating hearing aid personalisation. [ABSTRACT FROM AUTHOR]

Automatic differentiation is a popular technique for computing derivatives of computer programs. While automatic differentiation has been successfully used in countless engineering, science, and machine learning applications, it can sometimes nevertheless produce surprising results. In this paper, we categorize problematic usages of automatic differentiation, and illustrate each category with examples such as chaos, time‐averages, discretizations, fixed‐point loops, lookup tables, linear solvers, and probabilistic programs, in the hope that readers may more easily avoid or detect such pitfalls. We also review debugging techniques and their effectiveness in these situations. This article is categorized under:Technologies > Machine Learning [ABSTRACT FROM AUTHOR]

This paper examines the problem of synthesizing a two-loop control system for a three-phase voltage source inverter designed for an autonomous power supply system. This type of DC/AC converters is an essential part of power supply systems with renewable energy sources as primary sources, and, accordingly, must corresponds to the requirements for the electric power quality. These requirements are met through circuit design solutions, as well as through the construction of high-precision automatic control systems. The system integrates: a hydrogen electrolyzer, a hydrogen storage system, and a hydrogen fuel cell. The paper proposes a method for calculating the parameters of PI controllers with resonant components, based on the time-scale separation method and allowing independent adjustment of the controller components. The introduction of a resonant component ensures high tracking accuracy for a desired main voltage harmonic and selective suppression of external harmonic influences. The proposed approach allows to synthesize an automatic control system that significantly improves the quality of the voltage inverter output parameters. [ABSTRACT FROM AUTHOR]

This paper describes a new method for measuring and characterizing electrical parameters of ceramic materials. In addition, a comparison of experimental results obtained by manual measurement and an application of a new method for characterizing doped BaTiO3 ceramics at different temperatures and frequencies was performed. An LCR meter and a programmable testing furnace were used to measure the material’s dielectric parameters as a function of temperature. When measuring the parameters using the manual method, the results were obtained by manual control of the LCR meter, that is, they were measured by manually setting the parameters. The results were compared with those obtained using a new method for automatic control of the LCR meter, for which an application was developed. This comparison shows that feature measurement can be done fully automatically. By measuring samples of Er/BaTiO3 doped ceramics, it was possible to compare the results obtained by both methods. Based on the analysis of the measured results, it can be concluded that the new method is more precise, and the results are easier to manipulate in further work. In addition, the new method enables greater reproducibility of measurements, compared to the application of the manual method, which allows one measurement to be performed during the day. The described new method provides the possibility to measure material characteristics without the presence of a human factor. [ABSTRACT FROM AUTHOR]

ABSTRACT Automatic modulation classification (AMC) is explained as accurately identifying a modulation of a received signal. AMC systems are a significant component of cognitive radio network (CRN) systems. It is difficult to perform modulation classification on an unsettled radio signal without any previous knowledge of the signal's properties. In this work, the deep learning‐aided AMC is suggested to solve the difficulties of the existing models. In the proposed approach, the modulation classification is attained by performing two steps: (a) data collection and (b) classification. Initially, the required data related to the cognitive environment is collected from online resources. Later, the garnered data are passed to the classification phase. The AMC is performed by the adaptive and dilated hybrid network (ADHN), which is the combination of a temporal convolution network (TCN) and a gated recurrent unit (GRU). The ADHN accurately classifies the modulation even in a noisy environment. The classification performance of the ADHN is further boosted by tuning the parameters of this network via the enriched remora optimization algorithm (EROA). This proposed modulation classification model is suitable for various channels. The comparative validation is performed to ensure the usefulness of the designed system via several measures. By experimental analysis, the proposed system acquires the high value of accuracy, precision, and f1‐score by 94.2, 80.2, and 86.7, respectively, when compared with classical approaches. In addition to this, other metrics are considered and obtained with more true value and less false value. Thus, it ensures the effectiveness of classifying the modulation types in CRNs. [ABSTRACT FROM AUTHOR]

To study the influence of speed factors on the stability of a tractor automatic navigation system, combined with the neural network control theory, the authors proposed a dual-objective joint sliding mode control method based on lateral position deviation and heading angle deviation, using a back propagation neural network to establish a two-wheel tractor-path dynamics model and a straight-line path tracking deviation model. The overall system simulation was carried out using Matlab/Simulink, and the reliability of the control method was verified. The experimental results showed that when the tractor was tracked with the automatic control of a linear path under the condition of variable speed, the maximum deviation of the lateral position deviation was 12.7 cm, and the average absolute deviation was kept within 4.88 cm; the maximum deviation of the heading angle deviation was 5°, and the average absolute deviation was kept within 2°; the maximum value of the actual rotation angle was 3.13°, and the standard deviation of the fluctuation was within 0.84°. Under the conditions of constant speed and variable speed, using the joint sliding mode control method designed by the authors, the dual-objective joint control of lateral position deviation and heading angle deviation could be realized, the controlled overshoot was small, the controlled deviation was small after reaching a stable state, and the adaptability to speed factors was strong, which basically could meet the accuracy requirements of farmland operations. [ABSTRACT FROM AUTHOR]

Machine-learning potentials provide computationally efficient and accurate approximations of the Born–Oppenheimer potential energy surface. This potential determines many materials properties and simulation techniques usually require its gradients, in particular forces and stress for molecular dynamics, and heat flux for thermal transport properties. Recently developed potentials feature high body order and can include equivariant semi-local interactions through message-passing mechanisms. Due to their complex functional forms, they rely on automatic differentiation (AD), overcoming the need for manual implementations or finite-difference schemes to evaluate gradients. This study discusses how to use AD to efficiently obtain forces, stress, and heat flux for such potentials, and provides a model-independent implementation. The method is tested on the Lennard-Jones potential, and then applied to predict cohesive properties and thermal conductivity of tin selenide using an equivariant message-passing neural network potential. [ABSTRACT FROM AUTHOR]

Automatic identification of modal frequencies can be used to directly estimate the real-time tension force of bridge cables and provide early damage alarming. However, a large amount of abnormal monitoring data usually exists in structural health monitoring (SHM) systems. Abnormal monitoring data may lead to faulty results of modal frequency identification and incorrect cable tension force estimation. Then, false or missing alarming of cable damage may arise. An automatic identification method of bridge cable modal frequencies under the influence of abnormal monitoring data is proposed in this study. The peak picking (PP) method is used to automatically obtain the original identification results of cable modal frequencies. To remove faulty frequency identification results, a multidimensional density-based clustering model is established. The cable acceleration data of the Waitan cable-stayed bridge are used to verify the accuracy of the proposed method. The influence of various abnormal monitoring data on frequency identification is investigated, and the accuracy of multidimensional clustering models is verified. The results show that abnormal monitoring data have a harmful influence on automatic modal frequency identification for bridge cables. The accuracy of the multidimensional clustering model for faulty frequency identification results is more than 99%. After removing the faulty frequency identification results, the correlation between the cable modal frequencies and environmental temperature becomes clearer and more reasonable. [ABSTRACT FROM AUTHOR]

Background/Objectives: Radial artery tracking (RAT) in the short-axis view is a pivotal step for ultrasound-guided radial artery catheterization (RAC), which is widely employed in various clinical settings. To eliminate disparities and lay the foundations for automated procedures, a pilot study was conducted to explore the feasibility of U-Net and its variants in automatic RAT. Methods: Approved by the institutional ethics committee, patients as potential RAC candidates were enrolled, and the radial arteries were continuously scanned by B-mode ultrasonography. All acquired videos were processed into standardized images, and randomly divided into training, validation, and test sets in an 8:1:1 ratio. Deep learning models, including U-Net and its variants, such as Attention U-Net, UNet++, Res-UNet, TransUNet, and UNeXt, were utilized for automatic RAT. The performance of the deep learning architectures was assessed using loss functions, dice similarity coefficient (DSC), and Jaccard similarity coefficient (JSC). Performance differences were analyzed using the Kruskal–Wallis test. Results: The independent datasets comprised 7233 images extracted from 178 videos of 135 patients (53.3% women; mean age: 41.6 years). Consistent convergence of loss functions between the training and validation sets was achieved for all models except Attention U-Net. Res-UNet emerged as the optimal architecture in terms of DSC and JSC (93.14% and 87.93%), indicating a significant improvement compared to U-Net (91.79% vs. 86.19%, p < 0.05) and Attention U-Net (91.20% vs. 85.02%, p < 0.05). Conclusions: This pilot study validates the feasibility of U-Net and its variants in automatic RAT, highlighting the predominant performance of Res-UNet among the evaluated architectures. [ABSTRACT FROM AUTHOR]

Cirque landforms, which are distinctive features of cold regions, play a pivotal role in understanding environmental changes and climate research. Currently, machine learning-based automatic glacier recognition models have significantly improved recognition speed. However, these models rely solely on elevation data from Digital Elevation Model (DEM), neglecting other topographic factors that influence cirque identification, such as slope and aspect. This study integrates DEM, slope, and aspect data through pseudo-color fusion technology and utilizes the RetinaNet network to achieve automatic cirque recognition. The pseudo-color fusion image-based recognition model outperforms methods solely relying on DEM, achieving mAP (Intersection over Union: IoU @0.5) of 82.3 %. In practical applications, the pseudo-color fusion image-based recognition model demonstrates balanced performance in identifying different categories of cirques, particularly excelling in recognizing intricate terrains and irregularly shaped cirques. [ABSTRACT FROM AUTHOR]

In this study, an automatic control technology for liquid-hydrogen zero boil-off was developed to improve the long-term storage capability of small-scale liquefaction systems. An innovative system capable of monitoring both the pressure within the heat pipe and the pressure within the LH 2 vessel was developed. This innovative approach ensures continuous efficiency and reduces downtime, significantly improving the overall performance of the hydrogen liquefaction process. The optimal conditions for automatic control were obtained through a cycle test in ON/OFF mode. In addition, experimental investigations were conducted sequentially to investigate the effect of the water level of liquid hydrogen such as 1st section, 2nd section and 3rd section of LH 2. As a result, these experiments successfully demonstrated zero boil-off of liquid hydrogen for long-term storage in the vessel, corresponding to a constant value in the liquid level meter in each section. The velocity of stabilization from the temperature and pressure perspectives in the vessel increased, and the entire cryocooler cycle time was significantly influenced by the amount of liquid hydrogen produced. • An automatic control technology for liquid-hydrogen zero boil-off was developed. • The pressure data on the liquid hydrogen vessel and heat exchanger were determined. • The effect of the water level of liquid hydrogen was investigated. • Experiments demonstrated zero boil-off of liquid hydrogen for long-term storage. • The entire cryocooler cycle time was influenced by the amount of liquid hydrogen. [ABSTRACT FROM AUTHOR]

The surge in text data has driven extensive research into developing diverse automatic summarization approaches to effectively handle vast textual information. There are several reviews on this topic, yet no large‐scale analysis based on quantitative approaches has been conducted. To provide a comprehensive overview of the field, this study conducted a bibliometric analysis of 3108 papers published from 2010 to 2022, focusing on automatic summarization research regarding topics and trends, top sources, countries/regions, institutions, researchers, and scientific collaborations. We have identified the following trends. First, the number of papers has experienced 65% growth, with the majority being published in computer science conferences. Second, Asian countries and institutions, notably China and India, actively engage in this field and demonstrate a strong inclination toward inter‐regional international collaboration, contributing to more than 24% and 20% of the output, respectively. Third, researchers show a high level of interest in multihead and attention mechanisms, graph‐based semantic analysis, and topic modeling and clustering techniques, with each topic having a prevalence of over 10%. Finally, scholars have been increasingly interested in self‐supervised and zero/few‐shot learning, multihead and attention mechanisms, and temporal analysis and event detection. This study is valuable when it comes to enhancing scholars' and practitioners' understanding of the current hotspots and future directions in automatic summarization. This article is categorized under:Algorithmic Development > Text Mining [ABSTRACT FROM AUTHOR]

The purpose of the research was to develop an automatic in-field weather station specifically for use in agricultural environments. The station was set up for use in a variety of weather conditions and research plant species. To present the data collected, a smart system named Met4Agriculture was developed. Five sensors were included in the solar powered in-field weather station design: light intensity, temperature, relative humidity, rainfall, and soil moisture measured at a depth of 10 centimeters. The Thai Meteorological Department calibrated sensors for temperature, relative humidity, rainfall, and soil moisture; Silpakorn University was responsible for the calibration of the light intensity sensor. The calibration results showed that temperature and relative humidity sensors had an accuracy deviation of approximately 1 °C and 1-3%RH, respectively. The variance of the rainfall sensor was ± 3 mm. The variance of the rainfall sensor was + 3 mm. The light intensity sensor had a deviation 1.17 times less than the actual measured value. The soil moisture sensor's measured moisture level trended in the same direction as the calculated soil moisture level, with a coefficient of correlation between the two values close to 1. Additionally, a program was designed to run on the LILYGO system and develop the Met4Agriculture program. The system could connect data from the IoT NΕΤΡΙΕ system to display the measurement data from the station. The program referenced the station's measurement points using the ID and token from the NETPIE system and further developed it into the Met4Agriculture Application, which is compatible with all platforms. It can display historical data for analysis of seasonal changes and set thresholds for each parameter that impacts agricultural crop performance during different seasons. This research outcome enables the development of a low-cost prototype for an automatic in-field weather station monitoring system for agricultural field. The prototype system shows statistically significant measurement accuracy comparable to that of standard instruments. The developed stations were installed in all five regions in Thailand, this system can continuously collect, transmit, and display real-time monitoring data through a web application and real-time mobile applications, compatible with both iOS and Android operating systems. [ABSTRACT FROM AUTHOR]

The ballast control of a floating dock mainly relies on manual operations, which can be time-consuming and requires skilled workers. This study proposes an automatic ballast control system for floating docks, which improves operational efficiency and safety during the vessel docking process. A numerical model is developed to simulate the dynamic process of the floating dock's operations, which includes a six degrees-of-freedom (6DOF) model, a hydrostatic force model, a hydrodynamic force model, and a hydraulic model. The hydrostatic force model is developed using the Archimedes law and a strip theory along the longitudinal direction. The hydrodynamic model is made based on the effects of added mass and dynamic damping. The hydraulic model is proposed to deal with the hydraulic calculation of the ballast water system. The present automatic ballast control is designed based on a modified proportional controller (P-controller) to control the valve opening angles when the pitch or roll angles are larger than the corresponding threshold values. Without using controllers, the roll angles of the dock can reach 8.9 deg and 13 deg during the ballasting and de-ballasting operations, respectively. The present modified P-controller with optimized control parameters can stabilize the dock during the de-ballasting operation and keep the maximum pitch and roll angles no larger than 0.016 deg and 0.0783 deg, respectively. During the ballasting operation with the same control parameters, the roll and pitch are below 0.0604 deg and 0.0145 deg, respectively. The present automatic control will be further implemented in the vessel docking cases and can significantly improve the stability of the dock. [ABSTRACT FROM AUTHOR]

An automatic identification system (AIS) is often installed on merchant ships and fishing boats to prevent collisions and ensure safe navigation. The location information of ships transmitted from AIS equipment can help maritime traffic control prevent accidents. The southern coast of Korea comprises a complex coastline with numerous fishing boats and transit vessels. In particular, the Tongyeong and Geoje Islands include high-altitude mountains and islands, resulting in several radio-shaded areas where AIS signals cannot be received, owing to geographical effects. However, only a few studies have explored this region and performed practical experiments on the reception status of AIS locations in radio-shaded areas. In this study, we performed an experiment in the Geoje Island Sea on the southern coast to analyze the impact of high terrain on the reception rate and status of automatic identification devices. Two identical pieces of AIS equipment were installed to generate multiple radio waves, and the location data transmitted via different antennae were compared. The experimental analysis forms the basis for identifying the exact location of ships in the event of maritime accidents, facilitating rapid rescue. Moreover, the accuracy of the location transmitted by the AIS equipment can aid in detecting the cause of accidents. [ABSTRACT FROM AUTHOR]

This article highlights the importance of Air Traffic Control in ensuring the security and efficiency of international aviation operations. Traditional RADAR systems face limitations in power consumption, coverage, and accuracy, leading to the development of innovative solutions like the Automatic Dependent Surveillance-Broadcast (ADS-B) system. However, the open nature of ADS-B's unencrypted signals poses security challenges, allowing attackers to intercept and manipulate data. To address this, the paper emphasizes the need for symmetric authentication, which uses shared secret keys through schemes like Message Authentication Codes and block cipher encryption. The choice between these methods depends on factors such as security, performance, and usability. Symmetric authentication is a practical solution, requiring robust shared keys and careful encryption algorithm selection. The research underscores the need for secure ADS-B systems and the trade-offs required in designing authentication schemes. [ABSTRACT FROM AUTHOR]

Purpose: This study examines the effectiveness of automatic speech recognition (ASR) for individuals with speech disorders, addressing the gap in performance between read and conversational ASR. We analyze the factors influencing this disparity and the effect of speech mode-specific training on ASR accuracy. Method: Recordings of read and conversational speech from 27 individuals with various speech disorders were analyzed using both (a) one speaker-independent ASR system trained and optimized for typical speech and (b) multiple ASR models that were personalized to the speech of the participants with disordered speech. Word error rates were calculated for each speech model, read versus conversational, and subject. Linear mixed-effects models were used to assess the impact of speech mode and disorder severity on ASR accuracy. We investigated nine variables, classified as technical, linguistic, or speech impairment factors, for their potential influence on the performance gap. Results: We found a significant performance gap between read and conversational speech in both personalized and unadapted ASR models. Speech impairment severity notably impacted recognition accuracy in unadapted models for both speech modes and in personalized models for read speech. Linguistic attributes of utterances were the most influential on accuracy, though atypical speech characteristics also played a role. Including conversational speech samples in model training notably improved recognition accuracy. Conclusions: We observed a significant performance gap in ASR accuracy between read and conversational speech for individuals with speech disorders. This gap was largely due to the linguistic complexity and unique characteristics of speech disorders in conversational speech. Training personalized ASR models using conversational speech significantly improved recognition accuracy, demonstrating the importance of domain-specific training and highlighting the need for further research into ASR systems capable of handling disordered conversational speech effectively. [ABSTRACT FROM AUTHOR]

Background: Automatic speech recognition (ASR) can potentially help older adults and people with disabilities reduce their dependence on others and increase their participation in society. However, maxillectomy patients with reduced speech intelligibility may encounter some problems using such technologies. Objectives: To investigate the accuracy of three commonly used ASR platforms when used by Japanese maxillectomy patients with and without their obturator placed. Methods: Speech samples were obtained from 29 maxillectomy patients with and without their obturator and 17 healthy volunteers. The samples were input into three speaker‐independent speech recognition platforms and the transcribed text was compared with the original text to calculate the syllable error rate (SER). All participants also completed a conventional speech intelligibility test to grade their speech using Taguchi's method. A comprehensive articulation assessment of patients without their obturator was also performed. Results: Significant differences in SER were observed between healthy and maxillectomy groups. Maxillectomy patients with an obturator showed a significant negative correlation between speech intelligibility scores and SER. However, for those without an obturator, no significant correlations were observed. Furthermore, for maxillectomy patients without an obturator, significant differences were found between syllables grouped by vowels. Syllables containing /i/, /u/ and /e/ exhibited higher error rates compared to those containing /a/ and /o/. Additionally, significant differences were observed when syllables were grouped by consonant place of articulation and manner of articulation. Conclusion: The three platforms performed well for healthy volunteers and maxillectomy patients with their obturator, but the SER for maxillectomy patients without their obturator was high, rendering the platforms unusable. System improvement is needed to increase accuracy for maxillectomy patients. [ABSTRACT FROM AUTHOR]

In the 1960s, meteorological measurements worldwide were still being made at traditional ‘met sites’, many of which are still in use today. This article describes the development at the Institute of Hydrology of what were to become among the first modern automatic weather stations. In this sense, they are of historic interest. [ABSTRACT FROM AUTHOR]

Landslides are the most serious geological disaster in our country, causing economic losses. Because they go undetected, a large number of landslides that have caused disasters are not in the catalogue. At present, Interferometric Synthetic Aperture Radar (InSAR) has been widely used in the identification of landslides. However, it is time-consuming, inefficient, etc., to survey landslides throughout our large country. In the context of massive SAR data, this problem is more obvious. Therefore, based on the current technique of using differential interferogram phase gradient stacking to avoid phase unwrapping errors, a landslide phase gradient dataset has been constructed. To validate the dataset's effectiveness and applicability, deep learning methods were introduced, applying the dataset to four networks: U-Net, Attention-Unet, Bisenet v2, and Deeplab v3. The results indicate that the phase gradient dataset performs well across different models, with the Attention-Unet network demonstrating the best performance. Specifically, the precision, recall, and accuracy on the test dataset were 0.8771, 0.8712, and 0.9834, respectively, and the accuracy on the validation dataset was 0.8523. Finally, in this paper, the model is applied to landslide identification in Gansu Province, China, during 2022-2023, and a total of 1882 landslides are found. These landslides are mainly concentrated in the south of Gansu Province, where the terrain is relatively undulating. The results show that this method can quickly and accurately realize landslide automatic identification in a wide area and provide technical support for large-scale landslide disaster surveys. [ABSTRACT FROM AUTHOR]

Shape optimization approaches to inverse design offer low‐dimensional, physically‐guided parameterizations of structures by representing them as combinations of primitives. However, on fixed grids, computing the gradient of a user objective via the adjoint variables method requires a product of forward/adjoint field solutions and the Jacobian of the simulation material distribution with respect to the structural shape parameters. Shape parameters often perturb global parts of the simulation grid resulting in many non‐zero Jacobian entries. These are often computed by finite‐difference (FD) in practice, and hence can be non‐trivial. In this work, the gradient calculation is accelerated by invoking automatic differentiation (AD) in instantiations of structural material distributions, enabled by the development of extensible differentiable feature‐mappings from parameters to primitives and differentiable effective logic operations (denoted AutoDiffGeo or ADG). ADG can also be used to accelerate FD‐based shape optimization by efficient boundary selection. AD‐enhanced shape optimization is demonstrated using three integrated photonic examples: a blazed grating coupler, a waveguide transition taper, and a polarization‐splitting grating coupler. The accelerations of the gradient calculation by AD relative to FD with boundary selection exceed 10×$\times$, resulting in total optimization wall time accelerations of 1.4×$1.4\times$–3.8×$3.8\times$ on the same hardware with no compromise to device figure‐of‐merit. [ABSTRACT FROM AUTHOR]

The fast automatic differentiation technique and its advantages as applied to the numerical solution of optimal control problems are described. Results obtained by applying this technique to the optimal control of thermal processes with phase transitions are presented. [ABSTRACT FROM AUTHOR]

This study presents a model of automatic speech recognition (ASR) that is designed to diagnose pronunciation issues in children with speech sound disorders (SSDs) to replace manual transcriptions in clinical procedures. Because ASR models trained for general purposes mainly predict input speech into standard spelling words, well-known high-performance ASR models are not suitable for evaluating pronunciation in children with SSDs. We fine-tuned the wav2vec2.0 XLS-R model to recognise words as they are pronounced by children, rather than converting the speech into their standard spelling words. The model was fine-tuned with a speech dataset of 137 children with SSDs pronouncing 73 Korean words that are selected for actual clinical diagnosis. The model’s Phoneme Error Rate (PER) was only 10% when its predictions of children’s pronunciations were compared to human annotations of pronunciations as heard. In contrast, despite its robust performance on general tasks, the state-of-the-art ASR model Whisper showed limitations in recognising the speech of children with SSDs, with a PER of approximately 50%. While the model still requires improvement in terms of the recognition of unclear pronunciation, this study demonstrates that ASR models can streamline complex pronunciation error diagnostic procedures in clinical fields. [ABSTRACT FROM AUTHOR]

Speaking skills generally receive little attention in traditional English as a Foreign Language (EFL) classrooms, and this is especially the case in secondary education in Indonesia. A vocabulary deficit and poor pronunciation skills hinder learners in their efforts to improve speaking proficiency. In the present study, we investigated the effects of using two language learning websites, I Love Indonesia (ILI) and NovoLearning (NOVO). These websites are equipped with Automatic Speech Recognition (ASR) technology, with each website providing different types of immediate feedback. We measured written receptive and productive vocabulary knowledge of 40 target words before and after the intervention in which 146 students practiced with these two ASR-based websites, and compared it to that of a control group (n = 86). The ASR-based websites successfully helped students enhance their receptive vocabulary. Twenty-four students participated in a spoken pre-and post-test pronouncing the same 40 target words. We successfully developed an approach to measure pronunciation skills which showed that the treatment groups outperformed the control group. Our results indicate that our technology is successful in improving vocabulary and pronunciation skills. [ABSTRACT FROM AUTHOR]

AbstractObjectiveDesignStudy sampleResultsConclusionsOne proposed method to improve sound localisation for bilateral cochlear implant (BiCI) users is to synchronise the automatic gain control (AGC) of both audio processors. In this study we tested whether AGC synchronisation in a dual-loop front-end processing scheme with a 3:1 compression ratio improves sound localisation acuity.Source identification in the frontal hemifield was tested in in an anechoic chamber as a function of (roving) presentation level. Three different methods of AGC synchronisation were compared to the standard unsynchronised approach. Both root mean square error (RMSE) and signed bias were calculated to evaluate sound localisation in the horizontal plane.Six BiCI users.None of the three AGC synchronisation methods yielded significant improvements in either localisation error or bias, neither across presentation levels nor for individual presentation levels. For synchronised AGC, the pooled mean (standard deviation) localisation error of the three synchronisation methods was 24.7 (5.8) degrees RMSE, for unsynchronised AGC it was 27.4 (7.5) degrees. The localisation bias was 5.1 (5.5) degrees for synchronised AGC and 5.0 (3.8) for unsynchronised.These findings do not support the hypothesis that the tested AGC synchronisation configurations improves localisation acuity in bilateral users of MED-EL cochlear implants. [ABSTRACT FROM AUTHOR]

With timeliness and efficiency being critical in the aviation maintenance industry, the need has been growing for smart technological solutions that help in optimizing and streamlining the different underlying tasks (Bergkvist & Sabbagh, 2021). One such task is the technical documentation of the performed maintenance operations (Siyaev & Jo, 2021a). Instead of paper-based documentation, voice tools that transcribe spoken logbook entries allow technicians to document their work right away in a hands-free and time efficient manner. However, an accurate automatic speech recognition (ASR) model requires large training corpora (Siyaev & Jo, 2021a), which are lacking in the domain of aviation maintenance. In addition, ASR models which are trained on huge corpora in standard English, perform poorly in such a technical domain with non-standard terminology (Siyaev & Jo, 2021b). Hence, this study investigates the extent to which fine-tuning an ASR model, pre-trained on standard English corpora, on limited in-domain data improves its recognition performance in the technical domain of aviation maintenance. We present a case study on one such pre-trained ASR model, wav2vec 2.0 (Baevski et al., 2020). Results show that fine-tuning the model on a limited anonymized dataset of maintenance logbook entries significantly reduces its error rates when tested on not only an anonymized in-domain dataset, but also a non-anonymized one. This suggests that any available aviation maintenance logbooks, even if anonymized for privacy, can be used to fine-tune general-purpose ASR models, and enhance their in-domain performance. Lastly, an analysis on the influence of voice characteristics on model performance stresses the need for balanced datasets representative of the population of aviation maintenance technicians. [ABSTRACT FROM AUTHOR]

This study examines the effect of real-time subtitles generated by automatic speech recognition (ASR) technology on interpreting accuracy and interpreters' cognitive load. Multiple measurements — including interpreting accuracy, the NASA-TLX for subjective ratings of cognitive load, eye-tracking and theta power as indicated by EEG recordings — were applied. Twenty-three professional simultaneous interpreters worked with a video recording of a speech presented in five conditions: a baseline without subtitles and then with subtitles of varying levels of precision (100%, 95%, 90% and 80%). The results reveal that the presence of subtitles significantly improved interpreting accuracy, with a suggested optimal precision rate of 90% or higher. The interpreters looked more at the subtitles, regardless of their level of precision, than the speaker. Contrary to our predictions, the presence of subtitles decreased, rather than increased, the cognitive load (although this outcome was shown by the EEG data only and not by the self-reported data). We conclude that the cognitive cost of processing subtitles as an additional information channel is offset by the cognitive gain achieved through visual prompting. The study highlights a complex effect of subtitles on interpreting, with such factors as subtitle presence and precision modulating the interpreters' cognitive load in such a workflow. [ABSTRACT FROM AUTHOR]

This paper presents an innovative deep-learning model for Automatic Modulation Classification (AMC) in wireless communication systems. The proposed architecture integrates Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, augmented by a Gaussian noise layer to mitigate overfitting. The integration of both networks seeks to enhance classification accuracy and performance by leveraging the unique capabilities of CNNs and LSTMs in capturing spatial and temporal features, respectively. The model is expected to distinguish between eight digital and two analog modulation modes. Experimental evaluation on the RadioML2016.10b dataset demonstrates a peak recognition accuracy of 93.2% at 18 dB SNR. Comparative analyses validate the superior performance of the proposed architecture. The Gaussian noise layer contributes significantly to a 3% performance improvement at 18 dB SNR. The model achieves recognition accuracy exceeding 96% for most modulation modes, highlighting its robustness. Finally, computational complexity analysis underscores the efficiency of the proposed architecture, reinforcing its practical viability. [ABSTRACT FROM AUTHOR]

Bridge information models are essential for bridge inspection, assessment, and management. LiDAR technology, particularly UAV LiDAR, offers a cost‐effective means to capture dense and accurate 3D coordinates of a bridge's surface. However, the structure of large‐scale bridges is complex, and existing commercial software still demands substantial manual effort to segment the components when constructing bridge information models for large‐scale bridges. This study introduces a novel approach to automatically segment the components of a long‐span rail‐and‐road cable‐stayed bridge from the entire point cloud obtained through UAV LiDAR. In this proposed approach, the geometric and topological constraints of various bridge components are thoroughly examined, and a combination of the coarse‐to‐fine concept and top‐down strategy is employed. The key structural elements, including piers, cable towers, wind fairing plate, stay‐cable, main truss, railway surfaces, and deck surfaces, are identified and segmented. The proposed methodology achieves an average accuracy of over 96% at the point level validated using datasets acquired by UAV LiDAR. [ABSTRACT FROM AUTHOR]

Purpose: This work introduces updated transcripts, disfluency annotations, and word timings for FluencyBank, which we refer to as FluencyBank Timestamped. This data set will enable the thorough analysis of how speech processing models (such as speech recognition and disfluency detection models) perform when evaluated with typical speech versus speech from people who stutter (PWS). Method: We update the FluencyBank data set, which includes audio recordings from adults who stutter, to explore the robustness of speech processing models. Our update (semi-automated with manual review) includes new transcripts with timestamps and disfluency labels corresponding to each token in the transcript. Our disfluency labels capture typical disfluencies (filled pauses, repetitions, revisions, and partial words), and we explore how speech model performance compares for Switchboard (typical speech) and FluencyBank Time-stamped. We present benchmarks for three speech tasks: intended speech recognition, text-based disfluency detection, and audio-based disfluency detection. For the first task, we evaluate how well Whisper performs for intended speech recognition (i.e., transcribing speech without disfluencies). For the next tasks, we evaluate how well a Bidirectional Embedding Representations from Transformers (BERT) text-based model and a Whisper audio-based model perform for disfluency detection. We select these models, BERT and Whisper, as they have shown high accuracies on a broad range of tasks in their language and audio domains, respectively. Results: For the transcription task, we calculate an intended speech word error rate (isWER) between the model's output and the speaker's intended speech (i.e., speech without disfluencies). We find isWER is comparable between Switchboard and FluencyBank Timestamped, but that Whisper transcribes filled pauses and partial words at higher rates in the latter data set. Within Fluency-Bank Timestamped, isWER increases with stuttering severity. For the disfluency detection tasks, we find the models detect filled pauses, revisions, and partial words relatively well in FluencyBank Timestamped, but performance drops substantially for repetitions because the models are unable to generalize to the different types of repetitions (e.g., multiple repetitions and sound repetitions) from PWS. We hope that FluencyBank Timestamped will allow researchers to explore closing performance gaps between typical speech and speech from PWS. Conclusions: Our analysis shows that there are gaps in speech recognition and disfluency detection performance between typical speech and speech from PWS. We hope that FluencyBank Timestamped will contribute to more advancements in training robust speech processing models. [ABSTRACT FROM AUTHOR]

Purpose: Amyotrophic lateral sclerosis (ALS) is a progressive, ultimately fatal disease causing progressive muscular weakness. Most people living with ALS (plwALS) experience dysarthria, eventually becoming unable to communicate using natural speech. Many wish to use speech for as long as possible. Personalized automated speech recognition (ASR) model technology, such as Google's Project Relate, is argued to better recognize speech with dysarthria, supporting maintenance of understanding through real-time captioning. The objectives of this study are how plwALS and communication partners use Relate in everyday conversation over a period of up to 12 months and how it may change with any decline in speech over time. Method: This study videoed interactions between three plwALS and communication partners. We assessed ASR caption accuracy and how well they preserved meaning. Conversation analysis was used to identify participants' own organizational practices in the accomplishment of interaction. Thematic analysis was used to understand better the participants' experiences of using ASR captions. Results: All plwALS reported lower-than-expected ASR accuracy when used in conversation and felt ASR captioning was only useful in certain contexts. All participants liked the concept of live captioning and were hopeful that future improvements to ASR accuracy may support their communication in everyday life. Conclusions: Training is needed on best practices for customization and practical use of ASR technology and for the limitations of ASR in conversational settings. Support is needed for those less confident with technology and to reduce misplaced allocation of ownership of captioning errors, risking negative effects on psychological well-being. [ABSTRACT FROM AUTHOR]

In recent years, automatic map matching has received great technical progress. However, when it comes to vague matching situations such as improper vocabulary use, there still lack reliable solutions. To handle the current gap, this paper proposes a novel automatic map matching method based on the hybrid computing framework of hidden Markov model (HMM) and conditional random field. First, the data filtering is completed by performing second-order transformation towards automatic matching conditions of HMM. Then, the data classification is completed using automatic data classification based on the conditional random field. After that, a hybrid computing framework with spatial elements and layer selection is built to generate map matching results. Finally, some simulation experiments are conducted for evaluation. For one thing, the trend of matching accuracy changes under specified conditions is basically the same as that of nonspecified conditions. The maximum difference in matching calculation values is about 3 times. However, once the vocabulary continues to increase, the difference in matching results between the two narrows to 10–20%. For the other thing, the matching accuracy of a specified state is higher than that of sending a specified state. While nonspecified fuzzy matching accuracy is about 3 times higher and nonspecified precision matching accuracy is about 50% higher. [ABSTRACT FROM AUTHOR]

A modified onboard neuro-fuzzy adaptive (NFA) helicopter turboshaft engine (HTE) automatic control system (ACS) is proposed, which is based on a circuit consisting of a research object, a regulator, an emulator, a compensator, and an observer unit. In this scheme, it is proposed to use the proposed AFNN six-layer hybrid neuro-fuzzy network (NFN) with Sugeno fuzzy inference and a Gaussian membership function for fuzzy variables, which makes it possible to reduce the HTE fuel consumption parameter transient process regulation time by 15.0 times compared with the use of a traditional system automatic control (clear control), 17.5 times compared with the use of a fuzzy ACS (fuzzy control), and 11.25 times compared with the use of a neuro-fuzzy reconfigured ACS based on an ANFIS five-layer hybrid NFN. By applying the Lyapunov method as a criterion, its system stability is proven at any time, with the exception of the initial time, since at the initial time the system is in an equilibrium state. The use of the six-layer ANFF NFN made it possible to reduce the I and II types of error in the HTE fuel consumption controlling task by 1.36...2.06 times compared with the five-layer ANFIS NFN. This work also proposes an AFNN six-layer hybrid NFN training algorithm, which, due to adaptive elements, allows one to change its parameters and settings in real time based on changing conditions or external influences and, as a result, achieve an accuracy of up to 99.98% in the HTE fuel consumption controlling task and reduce losses to 0.2%. [ABSTRACT FROM AUTHOR]

Background: Disease-vector mosquito monitoring is an essential prerequisite to optimize control interventions and evidence-based risk predictions. However, conventional entomological monitoring methods are labor- and time-consuming and do not allow high temporal/spatial resolution. In 2022, a novel system coupling an optical sensor with machine learning technologies (VECTRACK) proved effective in counting and identifying Aedes albopictus and Culex pipiens adult females and males. Here, we carried out the first extensive field evaluation of the VECTRACK system to assess: (i) whether the catching capacity of a commercial BG-Mosquitaire trap (BGM) for adult mosquito equipped with VECTRACK (BGM + VECT) was affected by the sensor; (ii) the accuracy of the VECTRACK algorithm in correctly classifying the target mosquito species genus and sex; (iii) Ae. albopictus capture rate of BGM with or without VECTRACK. Methods: The same experimental design was implemented in four areas in northern (Bergamo and Padua districts), central (Rome) and southern (Procida Island, Naples) Italy. In each area, three types of traps—one BGM, one BGM + VECT and the combination of four sticky traps (STs)—were rotated each 48 h in three different sites. Each sampling scheme was replicated three times/area. Collected mosquitoes were counted and identified by both the VECTRACK algorithm and operator-mediated morphological examination. The performance of the VECTRACK system was assessed by generalized linear mixed and linear regression models. Aedes albopictus capture rates of BGMs were calculated based on the known capture rate of ST. Results: A total of 3829 mosquitoes (90.2% Ae. albopictus) were captured in 18 collection-days/trap/site. BGM and BGM + VECT showed a similar performance in collecting target mosquitoes. Results show high correlation between visual and automatic identification methods (Spearman Ae. albopictus: females = 0.97; males = 0.89; P < 0.0001) and low count errors. Moreover, the results allowed quantifying the heterogeneous effectiveness associated with different trap types in collecting Ae. albopictus and predicting estimates of its absolute density. Conclusions: Obtained results strongly support the VECTRACK system as a powerful tool for mosquito monitoring and research, and its applicability over a range of ecological conditions, accounting for its high potential for continuous monitoring with minimal human effort. [ABSTRACT FROM AUTHOR]

Disturbances and uncertainties, ubiquitous in real‐world systems, complicate model‐based control design. A key challenge is developing a precise control strategy that takes all disturbances into account. The automatic dry‐type clutch is a nonlinear, uncertain, and continuous system, where influencing factors are regarded as total disturbances. An active disturbance rejection control (ADRC) framework with an adaptive extended state observer (AESO) is proposed. The AESO's key strength is its adaptability, adjusting gains over time to minimize state estimation errors and noise impacts. This optimizes disturbance estimation, enhancing closed‐loop system performance. Furthermore, we prove the stability of the AESO through the Lyapunov method and, based on this, set the controller parameters as a Hurwitz matrix to ensure the stability of the closed‐loop system. Experimental data from clutch control demonstrates the efficacy of this controller. It ensures clutch position accuracy with a mean value of 13.5255 mm and a root mean square error of 28.7368 mm. Notably, the AESO's observation errors are minimal, averaging at 0.4365 mm with a root mean square error of 0.7068 mm, even under conditions of uncertain dynamics and measurement noise. This performance surpasses traditional ADRC methods relying on linear extended state observers (ESO) and PI control, thereby advancing the practical applicability of ADRC. Finally, through simulations comparing with fuzzy ESO and nonlinear ESO, the results show that the proposed AESO can achieve the same effect as them. [ABSTRACT FROM AUTHOR]

Traditional surface damage detection algorithms for agricultural products cannot be applied to large-scale instances due to the difficulty of their implementation. Thus, research on automatic identification and classification algorithms for agricultural products surface damage based on deep learning has emerged. This study proposed a deep learning-based algorithm for automatic identification and classification of surface damage of agricultural products and compared it with six mainstream classification models in plantvillage in terms of generalization ability, training time, and amount of pre-training data. The results proved that the model proposed in this study was the best in all aspects and had the highest accuracy when the pre-training data reached 40,000. This study verified the superiority and generalization ability of the proposed model and provided a new solution and reference standard in the field of agricultural product surface damage detection. Further, the results provided valuable reference and inspiration for related research and practice. [ABSTRACT FROM AUTHOR]

Automatic modulation classification models based on deep learning models are at risk of being interfered by adversarial attacks. In an adversarial attack, the attacker causes the classification model to misclassify the received signal by adding carefully crafted adversarial interference to the transmitted signal. Based on the requirements of efficient computing and edge deployment, a lightweight automatic modulation classification model is proposed. Considering that the lightweight automatic modulation classification model is more susceptible to interference from adversarial attacks and that adversarial training of the lightweight auto‐modulation classification model fails to achieve the desired results, an adversarial attack defense system for the lightweight automatic modulation classification model is further proposed, which can enhance the robustness when subjected to adversarial attacks. The defense method aims to transfer the adversarial robustness from a trained large automatic modulation classification model to a lightweight model through the technique of adversarial robust distillation. The proposed method exhibits better adversarial robustness than current defense techniques in feature fusion based automatic modulation classification models in white box attack scenarios. [ABSTRACT FROM AUTHOR]

The characteristics of automatic dependent surveillance-broadcast (ADS-B) technology, such as unencrypted channels and open data protocols, make it extremely vulnerable to various intentional or unintentional spoofing and jamming, which seriously threatens air traffic safety. Traditional approaches do not update their high-bit data for a long time, which causes the situations of the same encoding results to have different aircraft positions and it is difficult to determine the real position of the aircraft and effectively detect ADS-B jamming. Therefore, it is urgent to propose more effective detection methods for ADS-B spoofing and jamming. First, the methods and principles of frequency shift calculation and automatic monitoring algorithm are introduced. Then, signal consistency detection and error robustness are analyzed, and a track-data-based detection model for abnormal ADS-B data are proposed. The track-data-based detection methods can use multi-point sampling judgment algorithm to realize the reception and processing of data bits. This method makes full use of the information of multiple sampling values of each data bit and the reference power value obtained in the preamble detection, and determines the data value and confidence value through the relationship between the two. The detection methods for ADS-B spoofing and jamming distinguishes the true and false signals and the authenticity of target positions by comparing the arrival time difference of each signal at the ground station. Finally, the research results show that, with the increase of signal to noise ratio, no matter the estimation based on virtual array translation or the estimation based on circular stationary signal, their estimation accuracies are improved with gradually increasing trends. [ABSTRACT FROM AUTHOR]

Automatic train operation (ATO) is a critical component of automatic train control (ATC) systems. The ATO automatically adjusts the speed of trains, ensuring the safety of trains and increasing the passenger capacity of urban railway networks. The traditional ATO models employ a linear approximation approach to fit the nonlinear control model of the train, such as proportional–integral–derivative (PID) or model predictive control (MPC). However, due to the complexity of actual train operations, these methods fail to accurately fit train models. Thus, they are unable to shorten the operational intervals of trains and meet the demand for improving system efficiency of ATC. In recent years, machine learning-based methods have been employed for deriving more accurate train control models. However, these methods require trains to offload massive amounts of data to central servers for centralized training, which is challenging in CBTC systems due to limited wireless bandwidth. In this paper, we propose using federated learning (FL)-based edge computing to train the ATO model without transmitting a large amount of data. The onboard computing devices are used to perform local federated learning when the train is idle. Additionally, we present a Tab-Transformer-based machine learning model for ATO policy to improve the prediction accuracy of federally trained models. Our extensive simulation results demonstrate that the proposed federated learning scheme improves the efficiency of model training, and the Tab-Transformer-based ATO model achieves better driving performance. [ABSTRACT FROM AUTHOR]

Investigating subsurface shear wave velocity (v s) structures using surface wave dispersion data involves minimizing a misfit function that is commonly solved through gradient-based optimization. Sensitivity kernels for model updates are commonly estimated using numerical differentiation, variational methods or implicit functions which however, may involve numerical instability and computational challenges when dealing with complex velocity models and large data sets. In this study, we propose a novel surface wave inversion framework in which error-free gradients are calculated by automatic differentiation (AD) and forward modelling is implemented by convenient computational graphs in the state-of-the-art deep learning framework. The AD-based inversion approach is first validated using two synthetic data sets. Then, the subsurface structures at three distinct locations, namely the Great Plains and the Long Beach in the US and Tong Zhou in China, are also derived using this method with seismic ambient noise data, which show nice consistency with those obtained using traditional methods. With the significantly improved computational efficiency, a great number of initial models can be inverted simultaneously to mitigate the impact of local minima and to estimate the uncertainty in the invert models. We have developed a new surface wave inversion package named ADsurf based on automatic differentiation and computational graphs in the deep learning framework, and its computational efficiency is also compared with the traditional finite-difference-based gradient estimation approach. While a great number of intriguing studies on the geophysical inverse problems have been conducted recently using deep learning for end-to-end mapping, the use of AD provided in the in the deep learning frameworks to assist and expedite the gradient computations are still underexploited in geophysics. Thus, it is expected that various geophysical inverse problems in many different areas beyond the surface wave inversion can also be tackled with this new paradigm in the future. [ABSTRACT FROM AUTHOR]

Prior research has associated spoken language use with depression, yet studies often involve small or non-clinical samples and face challenges in the manual transcription of speech. This paper aimed to automatically identify depression-related topics in speech recordings collected from clinical samples. The data included 3919 English free-response speech recordings collected via smartphones from 265 participants with a depression history. We transcribed speech recordings via automatic speech recognition (Whisper tool, OpenAI) and identified principal topics from transcriptions using a deep learning topic model (BERTopic). To identify depression risk topics and understand the context, we compared participants' depression severity and behavioral (extracted from wearable devices) and linguistic (extracted from transcribed texts) characteristics across identified topics. From the 29 topics identified, we identified 6 risk topics for depression: 'No Expectations', 'Sleep', 'Mental Therapy', 'Haircut', 'Studying', and 'Coursework'. Participants mentioning depression risk topics exhibited higher sleep variability, later sleep onset, and fewer daily steps and used fewer words, more negative language, and fewer leisure-related words in their speech recordings. Our findings were derived from a depressed cohort with a specific speech task, potentially limiting the generalizability to non-clinical populations or other speech tasks. Additionally, some topics had small sample sizes, necessitating further validation in larger datasets. This study demonstrates that specific speech topics can indicate depression severity. The employed data-driven workflow provides a practical approach for analyzing large-scale speech data collected from real-world settings. • Transcribed smartphone-collected speech recordings via automatic speech recognition • Utilized a deep learning topic model for identifying topics from transcriptions • Identified six risk topics associated with high depression symptom severity • Explored the contexts of topics using behavioral and linguistic characteristics [ABSTRACT FROM AUTHOR]

Automatic tracking of three-dimensional (3D) human motion pose has the potential to provide corresponding technical support in various fields. However, existing methods for tracking human motion pose suffer from significant errors, long tracking times and suboptimal tracking results. To address these issues, an automatic tracking method for 3D human motion pose using contrastive learning is proposed. By using the feature parameters of 3D human motion poses, threshold variation parameters of 3D human motion poses are computed. The golden section is introduced to transform the threshold variation parameters and extract the features of 3D human motion poses by comparing the feature parameters with the threshold of parameter variation. Under the supervision of contrastive learning, a constraint loss is added to the local–global deep supervision module of contrastive learning to extract local parameters of 3D human motion poses, combined with their local features. After normalizing the 3D human motion pose images, frame differences of the background image are calculated. By constructing an automatic tracking model for 3D human motion poses, automatic tracking of 3D human motion poses is achieved. Experimental results demonstrate that the highest tracking lag is 9%, there is no deviation in node tracking, the pixel contrast is maintained above 90% and only 6 sub-blocks have detail loss. This indicates that the proposed method effectively tracks 3D human motion poses, tracks all the nodes, achieves high accuracy in automatic tracking and produces good tracking results. [ABSTRACT FROM AUTHOR]

Automatically identifying the working status of transformers can promptly detect faults and abnormalities, which provides a scientific basis for maintenance and management. To response the challenges of complex data processing and low recognition accuracy in traditional transformer working state monitoring methods, an automatic identification method of transformer working state based on big data mining was proposed. First, by using the K-nearest neighbor improved fast density peak clustering algorithm, the transformer Supervisory Control and Data Acquisition (SCADA) big data were clustered to obtain transformer big data of normal and abnormal states. Then, combining the big data of transformers from the two categories, the Γ-type equivalent algorithm was used to estimate the transformer working parameters for each category. Finally, within the least-squares support vector machine, the transformer working parameters were input to output the automatic identification results of the transformer working status. Results show that the method proposed in this study can accurately cluster transformer SCADA big data and estimate transformer working parameters. The analysis of the relative error frequency distribution histogram show that the relative error of this method in automatically identifying the transformer working status is only ±0.02, which demonstrates high automatic identification accuracy. The automatic identification method proposed in this study provides technical support for the daily operation and maintenance of transformers. [ABSTRACT FROM AUTHOR]

In global optimization problems, diversification approaches are often necessary to overcome the convergence toward local optima. One approach is the multi-start method, where a set of different starting configurations are taken into account to designate the best local minimum returned by the multiple optimization procedures as the (possible) global optimum. Therefore, parallelization is crucial for multi-start. In this work, we present a new multi-start approach for gradient-based optimization methods that exploits the reverse Automatic Differentiation to perform efficiently. In particular, for each step, this Automatic Differentiation-based method is able to compute the N gradients of N optimization procedures extremely quickly, exploiting the implicit parallelization guaranteed by the computational graph representation of the multi-start problem. The practical advantages of the proposed method are illustrated by analyzing the time complexity from a theoretical point of view and showing numerical examples where the speed-up is between × 40 and × 100 , with respect to classic parallelization methods. Moreover, we show that our AD-based multi-start approach can be implemented by using tailored shallow Neural Networks, taking advantage of the built-in optimization procedures of the Deep Learning frameworks. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicle (UAV)-assisted communication based on automatic modulation classification (AMC) technology is considered an effective solution to improve the transmission efficiency of wireless communication systems, as it can adaptively select the most suitable modulation method according to the current communication environment. However, many existing deep learning (DL)-based AMC methods cannot be directly applied to UAV platform with limited computing power and storage space, because of the contradiction between accuracy and efficiency. This paper mainly studies the lightweight of DL-based AMC networks to improve adaptability in resource-constrained scenarios. To address this challenge, we propose an ultra-lightweight neural network (ULNN). This network incorporates a lightweight convolutional structure, attention mechanism, and cross-channel feature fusion technique. Additionally, we introduce data augmentation (DA) based on signal phase offsets during the model training process, aimed at improving the model's generalization ability and preventing overfitting. Through experimental validation on the public dataset RML2016.10 A, the ULNN we proposed achieves an average precision of 62.83% with only 8815 parameters and reaches a peak classification accuracy of 92.11% at SNR = 10 dB. The experimental results show that ULNN can achieve high recognition accuracy while keeping the model lightweight, and is suitable for UAV platform with limited resources. [ABSTRACT FROM AUTHOR]

The negative surface mass balance of glaciers and ice caps under a warming climate impacts local ecosystems, influencing the volume and timing of water flow in local catchments while also contributing to global sea level rise. Peripheral glaciers distinct to the Greenland ice sheet respond faster to climate change than the main ice sheet. Accurate assessment of surface mass balance depends on in situ observations of near-surface climate and ice ablation, but very few in situ observations of near-surface climate and ice ablation are freely available for Greenland's peripheral glaciers. The transect of three automated weather and ablation stations on the peripheral A. P. Olsen Ice Cap in northeast Greenland is an example of these much needed data. The transect has been monitored since 2008, and in 2022, the old weather and ablation stations were replaced by a new standardized setup. In order to ensure comparable data quality of the old and new monitoring station setups, it is necessary to re-evaluate the data collected between 2008 and 2022. This paper presents the fully reprocessed near-surface climate and ablation data from the A. P. Olsen Ice Cap transect from 2008 to 2022, with a focus on data quality and the usability in ice ablation process studies. The usability of the data is exemplified by the data in an energy balance melt model for two different years. We show that the inherent uncertainties in the data result in an accurate reproduction of ice ablation for just one of the two years. A transect of three automatic ablation and weather stations of this length is unique to Greenland's peripheral glaciers, and it has a broad scale of usage from input to climate reanalysis and detailed surface ablation studies. The dataset can be downloaded at 10.22008/FK2/X9X9GN. [ABSTRACT FROM AUTHOR]