Extraction Performance Research Articles

Evaluating The Performance of DWT-DCT Feature Extraction in Guitar Chord Recognition

This study presents advancements in audio signal processing techniques, specifically in enhancing the efficiency of guitar chord recognition. It is a continuation of the previous studies, which also aim at minimizing the feature extraction length with the intended performance. This study adopted two signal processing techniques that are common: Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) for use in the feature extraction method. By conducting a systematic evaluation of two key parameters: frame blocking length and wavelet filter selection, a significant achievement could be achieved. The recognition system managed to obtain chord recognition with an accuracy of up to 91.43%, by using a feature extraction length of only three, which brought about smaller representation than the previous studies. The outcome of this study will help improve the data processing, which can be applied in real time, in this case in Field Programmable Gate Array (FPGA)-based chord recognition systems. Keywords: chord recognition, Discrete Wavelet Transform, Discrete Cosine Transform, feature extraction

Deep Learning's Current and Future Applications in Predicting Vehicle Traffic Flow

With the acceleration of urbanization and the continuous growth of traffic flow, accurate prediction of highway traffic flow has become a key demand for optimizing traffic management and alleviating congestion. Traditional prediction methods are often inefficient in dealing with complex traffic patterns, while deep learning technology brings new solutions for traffic flow prediction with its powerful feature learning and pattern recognition capabilities. This paper reviews deep learning applications' current state and future trends in highway traffic flow prediction. It introduces deep learning technology and its advantages in traffic flow prediction. The paper then analyzes the specific applications of Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and their variants in data processing, feature extraction, and prediction performance. The challenges currently faced, such as data preprocessing and the capture of long-term dependencies, are also discussed, along with proposed solutions. Finally, the paper looks forward to the future development of deep learning in the field of traffic flow prediction, highlighting its importance in improving traffic management efficiency.

Robotics Classification of Domain Knowledge Based on a Knowledge Graph for Home Service Robot Applications

The representation and utilization of environmental information by service robots has become increasingly challenging. In order to solve the problems that the service robot platform has, such as high timeliness requirements for indoor environment recognition tasks and the small scale of indoor scene data, a method and model for rapid classification of household environment domain knowledge is proposed, which can achieve high recognition accuracy by using a small-scale indoor scene and tool dataset. This paper uses a knowledge graph to associate data for home service robots. The application requirements of knowledge graphs for home service robots are analyzed to establish a rule base for the system. A domain ontology of the home environment is constructed for use in the knowledge graph system, and the interior functional areas and functional tools are classified. This designed knowledge graph contributes to the state of the art by improving the accuracy and efficiency of service decision making. The lightweight network MobileNetV3 is used to pre-train the model, and a lightweight convolution method with good feature extraction performance is selected. This proposal adopts a combination of MobileNetV3 and transfer learning, integrating large-scale pre-training with fine-tuning for the home environment to address the challenge of limited data for home robots. The results show that the proposed model achieves higher recognition accuracy and recognition speed than other common methods, meeting the work requirements of service robots. With the Scene15 dataset, the proposed scheme has the highest recognition accuracy of 0.8815 and the fastest recognition speed of 63.11 microseconds per sheet.

Endowing deep eutectic solvent with neutral and ionizable perfluoroalkyl substances extracting capability: experimental study and ab initio molecular dynamics simulation

ABSTRACT The simultaneous separation of neutral and ionizable perfluoroalkyl substances (PFASs) is challenging. The traditional solvents used for this purpose are toxic and harmful. Deep eutectic solvents (DESs) as sustainable alternatives to toxic reagents have received considerable attention in current research. In this work, an array of complementary experimental (liquid–liquid equilibrium experiments) and computational (ab initio molecular dynamics, AIMD) techniques were utilized to explore the distinct distribution behaviors, complicated interaction network, and corresponding mechanisms from multidimensional temporal and spatial perspectives. The proposed bromophenylcyanide-based eutectic solvent systems exhibited good separation and extraction performance for both neutral and ionizable PFASs. The presence of a cyano group adjacent to bromine leads to increased polarization of the C–Br bond, resulting in increased strength of the Br···F contact, which provides selectivity to some extent. For fatty acids, their disruptive effect on eutectic systems is greater because of the relatively strong aggregation that affects DES structuring. For PFASs, different categories of target monomers are confined in the DES cavities and interact with the surrounding solvent through abundant interaction sites (such as electrostatic attraction, N-H···F-C, O-H···F-C, O-H···O-H, and C≡N···O-H hydrogen bonds) without significantly disrupting the solvent structural properties by reducing the molecular mobility.

Sorbent-Based Sampling With Two-Stage Trapping/Desorption Coupled to Comprehensive Two-Dimensional Gas Chromatography and Mass Spectrometry for Terpenoids Profiling in Cannabis.

Cannabis inflorescences represent an important source of many high-value bioactive specialized metabolites, among which the family of terpenes or terpenoids that are the largest classes of natural products known. Besides their biological activities either alone or synergistic with other terpenoids and/or cannabinoids, they are responsible for their distinctive flavour. In this study, we exploited the separation power and identification capabilities of comprehensive two-dimensional gas chromatography coupled to mass spectrometry (GC×GC-MS) for the profiling of terpenes and terpenoids in cannabis inflorescences. The dynamic headspace (DHS) used herein for the extraction was chosen for its sensitivity, portability, suitability, as well as its versatility of sampling various natural products, including plant raw materials and different plant parts. The enrichment method and the following desorption into the GC were developed and optimized on both standards and real samples considering different sorbent traps (i.e. Tenax-TA, Carbotrap T420, Carbotrap 202), and evaluating key performance values. Analyte coverage, recovery and response reproducibility were used for the evaluation of the best performing thermal desorption tube. Considering terpenoids profiling on cannabis inflorescences, satisfactory extraction performance was observed with both Tenax-TA and Carbotrap T420. However, Tenax-TA provided a wider analyte coverage beyond the class of terpenoids, thus can be better suited for non-targeted analysis. On the other hand, peak width, peak height, peak quality and resolution were considered for the optimization of the chromatographic process, and more specifically the injection process, demonstrating the benefit of a secondary trapping/desorption stage with a cryotrap. Finally, considering the final DHSE-TD-GC×GC-MS conditions, terpenes and terpenoids were profiled in real-world cannabis inflorescences, highlighting the differences among the chemovars.

Segmentation and Proportion Extraction of Crop, Crop Residues, and Soil Using Digital Images and Deep Learning

Conservation tillage involves covering the soil surface with crop residues after harvest, typically through reduced or no-tillage practices. This approach increases the soil organic matter, improves the soil structure, prevents erosion, reduces water loss, promotes microbial activity, and enhances root development. Therefore, accurate information on crop residue coverage is critical for monitoring the implementation of conservation tillage practices. This study collected “crop–crop residues–soil” images from wheat-soybean rotation fields using mobile phones to create calibration, validation, and independent validation datasets. We developed a deep learning model named crop–crop residue–soil segmentation network (CCRSNet) to enhance the performance of cropland “crop–crop residues–soil” image segmentation and proportion extraction. The model enhances the segmentation accuracy and proportion extraction by extracting and integrating shallow and deep image features and attention modules to capture multi-scale contextual information. Our findings indicated that (1) lightweight models outperformed deeper networks for “crop–crop residues–soil” image segmentation. When CCRSNet employed a deep network backbone (ResNet50), its feature extraction capability was inferior to that of lighter models (VGG16). (2) CCRSNet models that integrated shallow and deep features with attention modules achieved a high segmentation and proportion extraction performance. Using VGG16 as the backbone, CCRSNet achieved an mIoU of 92.73% and a PA of 96.23% in the independent validation dataset, surpassing traditional SVM and RF models. The RMSE for the proportion extraction accuracy ranged from 1.05% to 3.56%. These results demonstrate the potential of CCRSNet for the accurate, rapid, and low-cost detection of crop residue coverage. However, the generalizability and robustness of deep learning models depend on the diversity of calibration datasets. Further experiments across different regions and crops are required to validate this method’s accuracy and applicability for “crop–crop residues–soil” image segmentation and proportion extraction.

“Double vaccinated, 5G boosted!”: Learning Attitudes Towards COVID-19 Vaccination from Social Media

The sudden onset of the recently concluded COVID-19 pandemic has driven substantial progress in various scientific fields. One notable example is the comprehension of public vaccination attitudes and the timely monitoring of their fluctuations through social media platforms. This approach can serve as a cost-effective means to supplement surveys in gathering public vaccine hesitancy levels. In this paper, we propose a deep-learning framework leveraging textual posts on social media to extract and track users’ vaccination stances in near real-time. Compared to previous works, we integrate into the framework the recent posts of a user’s social network friends to collaboratively detect the user’s genuine attitude towards vaccination. Based on our annotated dataset from X (previously known as Twitter), the models instantiated from our framework can increase the performance of attitude extraction by up to 23% compared to the state-of-the-art text-only models. Using this framework, we successfully confirm the feasibility of using social media to track the evolution of vaccination attitudes in real life. In addition, we illustrate the generality of our framework in extracting other public opinions such as political ideology. We further show one practical use of our framework by validating the possibility of forecasting a user’s vaccine hesitancy changes with information perceived from social media.

Analysis of mechanical properties of cased telescoped ammunitions during firing

Abstract In order to solve the cartridge extraction problem of cased telescoped ammunitions, a study was conducted on the dynamic process of cartridge during launching. The finite element simulation model is established firstly to analyse the mechanical properties of the cartridge during launching. Then the radial residual deformation and cartridge extraction force of the cartridge after chamber pressure unloading were obtained. Meanwhile, the effect of different yield limits of cartridge materials on the extraction performance was studied. The results show that as the yield limit of the cartridge increases, the maximum radial residual deformation first increases and then gradually decreases after unloading the chamber pressure, and the extraction force of the cartridge also shows the same trend. When the yield limit of the cartridge is greater than 1040 MPa, the extraction force goes to zero after pressure unloading, and the cartridge can be easily extracted.

MSR70 Re-Training of the Artificial Intelligence Tool LiveRefTM: Improved Accuracy and Performance in Data Extraction

MSR70 Re-Training of the Artificial Intelligence Tool LiveRefTM: Improved Accuracy and Performance in Data Extraction

Diabetic Retinopathy Detection Model using Hybrid of U-Net and Vision Transformer Algorithms

Diabetic retinopathy is one of the leading causes of vision impairment noticed among individuals with prolonged diabetes. Early-stage detection is very crucial for its treatment. Now, we present a hybrid model which is a combination of U-Net algorithm used for image segmentation and Vision Transformer for classification. The total integration offers a robust model which helps in detecting various stages of diabetic retinopathy. We leverage the use of U-Net algorithm in image segmentation process to delineate the regions of interest in retinal images. Further, the outputs which are segmented are passed into Vision Transformer, which is enhanced by Efficient Net, which is used across various severity levels involved in Diabetic Retinopathy. The usage of transformer architecture helps improve feature extraction and classification performance which ensures that our model captures all patterns in retinal images. We have evaluated our model on APTOS Blindness detection dataset in which our model outperforms traditional convolutional neural networks-based models. Hence, the hybrid approach consisting of combination of both the algorithms demonstrates excellent robustness and generalization which offers a promising application for diabetic retinopathy screening, involving the potential to revolutionize early diagnosis in clinical settings.

Influence of sweep angle on leading edge vortex dynamics of a fully passive oscillating-plate hydrokinetic turbine prototype

The present work assesses the use of spanwise flow in enhancing the performance of the fully passive oscillating-plate hydrokinetic turbine prototype by relating the dynamics of the leading edge vortex (LEV) to the power extraction performance. Two-dimensional (2D) planar and three-dimensional (3D) tomographic particle image velocimetry were employed to obtain quantitative flow structure of oscillating-plates undergoing heave and pitch motion in uniform inflow at Reynolds number of 21 000. A plate with a 6° sweep angle and an unswept plate (control case) were considered in the present study. Dynamics of the LEV were investigated using the patterns of the phase-averaged vorticity during the oscillation cycle. The 3D vector fields provided insights into the spanwise variation of the vortical structure and the rate of deformation of the vortex, which was determined by calculating the deformation terms in the vorticity transport equations which are related to the stability of the vortex. The results show evidence of a delay in the shedding of the LEV and an increase in its stability in the case of the swept plate, compared to the control case, which in turn benefit the power extraction performance at high inflow velocities.

CFD modelling and experimental analysis of aromatic amine extraction in a flat sheet supported liquid membrane contactor

Supported liquid membranes (SLMs) using ionic liquids are effective for the extraction of aromatic amines. This experimental study employed a flat sheet SLM contactor with the ionic liquid trihexyltetradecylphosphonium bis(trifluoromethylsulfonyl)imide ([P6,6,6,14][N(Tf)2]) as the solvent to investigate the separation of α-methylbenzylamine (MBA) and 1-methyl-3-phenylpropylamine (MPPA) from isopropyl amine (IPA). A detailed process study was conducted to examine the effects of flow rate (5–10 L/h), feed concentration (0.5–2.5 g/L), and feed pH (9–11) on extraction performance. Under standard experimental conditions (10 L/h, 1.0 g/L, pH 10), MBA and MPPA demonstrated high solute fluxes of 2.39 and 5.47 g/(m2h), respectively, compared with IPA, which had a solute flux of 0.84 g/(m2h). However, after 24 h, the recoveries were relatively low, at 17.9 % for MBA, 32.6 % for MPPA, and 5.2 % for IPA. No significant velocity dependency was observed, with slight variations attributed to minor pH changes, while a linear flux increase was noted for higher feed concentrations. The feed pH had a significant impact on the extraction performance, with higher pH levels resulting in increased solute fluxes and recoveries. To complement the experimental results, computational fluid dynamics (CFD) simulations were employed using COMSOL Multiphysics 5.1. The model demonstrated satisfactory agreement across various conditions, but underestimated fluxes and recoveries at higher pH values. Consequently, a new mass transfer mechanism was proposed to explain the variations observed in the experimental results.

A Multifocal Graph-Based Neural Network Scheme for Topic Event Extraction

Event extraction is a long-standing and challenging task in natural language processing, and existing studies mainly focus on extracting events within sentences. However, a significant problem that has not been carefully investigated is whether an “event topic” can be identified to represent the main aspects of extracted events. This article formulates the “topic event” extraction problem, aiming to identify a representative event from extracted ones. Specifically, after defining the topic event, we develop a multifocal graph-based framework to handle the extraction task. To enrich the associations of events and their tokens, we construct four event graphs, including the event subgraph and three event-associated graphs (i.e., event dependency parsing graph, event organization graph, and event share token graph), that reflect the internal and external structures of events, respectively. Subsequently, we design a multi-attention event-graph neural network to capture these event graph structures and improve event subgraph embedding. Finally, the output embeddings in the last layer of each channel are concatenated and fed into a fully connected network for topic event recognition. Extensive experiments validate the effectiveness of our method, and the results confirm its superiority over state-of-the-art baselines. In-depth analyses explore the essential factors (e.g., graph structures, attentions, feature generation method, etc.) determining the extraction performance.

Exoskeleton Robot Based on Brain-Computer Interface

Brain-computer interface (BCI) technology represents a means of facilitating human-computer interaction. One of the most widely accepted paradigms of brain-computer interface is motor imagery, which enables the recognition of electroencephalogram (EEG) signals generated in a specific brain region by imagining the movement of a limb. Following the acquisition, preprocessing, feature processing, and signal classification of the EEG signals, the complex signals are accurately recognized. Therefore, by creating a control system that translates the recognized EEG signals into movement commands for the robot and transmits them to the robot, it is possible to control the robot's movements by motor imagery. The convolutional neural network is the most popular signal processing algorithm due to its high EEG recognition accuracy, excellent performance in feature extraction, and superior performance in end-to-end learning. The convolutional neural network is an optimal method for signal processing in robot control. This makes CNN an optimal choice for processing EEG signals in robot control, enhancing both the effectiveness and user experience of BCI systems by enabling more intuitive and responsive interactions with robotic devices.

Numerical research on geothermal energy extraction in backfilled mines by using the excellent heat transfer performance of loop heat pipe

Deep mine harbor substantial geothermal energy. Integrating a heat exchanger within the backfill body mitigates thermal hazards and facilitates the concurrent extraction of mineral and geothermal resources. Inspired by the capabilities of superior thermal conductivity, high heat transfer limit without additional energy consumption of loop heat pipe (LHP), a novel cemented paste backfill system coupled with an LHP heat exchanger (LHPHE-CPB) was developed to effectively improve the thermal conductivity of backfill body and enhance the extraction performance of geothermal energy in backfilled mines. The temperature evolutions of LHPHE-CPB system and the mechanisms of vapor-liquid phase transition and two-phase flow within LHP were numerically analyzed during the stages of heat storage and simultaneous heat storage/heat release. Orthogonal tests meticulously examined the effects of surrounding rock temperature, and the inlet temperature and flow rate of cooling water on the system's heat transfer performance. Optimal operating conditions for the system, in terms of reducing backfill body temperature, achieving favorable temperature differentials between the inlet and outlet cooling water, and enhancing the heat extraction capacity of system, were determined through range and variance analyses. This research establishes a theoretical foundation for the application of LHP in efficiently extracting geothermal energy from backfilled mines.

Extracting International Classification of Diseases Codes from Clinical Documentation using Large Language Models.

Large language models (LLMs) have shown promise in various professional fields, including medicine and law. However, their performance in highly specialized tasks, such as extracting ICD-10-CM codes from patient notes, remains underexplored. The primary objective was to evaluate and compare the performance of ICD-10-CM code extraction by different LLMs with that of human coder. We evaluated performance of six LLMs (GPT-3.5, GPT-4, Claude 2.1, Claude 3, Gemini Advanced, and Llama 2-70b) in extracting ICD-10-CM codes against human coder. We used deidentified inpatient notes from American Health Information Management Association Vlab authentic patient cases for this study. We calculated percent agreement and Cohen's kappa values to assess the agreement between LLMs and human coder. We then identified reasons for discrepancies in code extraction by LLMs in a 10% random subset. Among 50 inpatient notes, human coder extracted 165 unique ICD-10-CM codes. LLMs extracted significantly higher number of unique ICD-10-CM codes than human coder, with Llama 2-70b extracting most (658) and Gemini Advanced the least (221). GPT-4 achieved highest percent agreement with human coder at 15.2%, followed by Claude 3 (12.7%) and GPT-3.5 (12.4%). Cohen's kappa values indicated minimal to no agreement, ranging from -0.02 to 0.01. When focusing on primary diagnosis, Claude 3 achieved highest percent agreement (26%) and kappa value (0.25). Reasons for discrepancies in extraction of codes varied amongst LLMs and included extraction of codes for diagnoses not confirmed by providers (60% with GPT-4), extraction of non-specific codes (25% with GPT-3.5), extraction of codes for signs and symptoms despite presence of more specific diagnosis (22% with Claude-2.1) and hallucinations (35% with Claude-2.1). Current LLMs have poor performance in extraction of ICD-10-CM codes from inpatient notes when compared against the human coder.

Numerical investigation on heat extraction performance of supercritical CO2 in depleted oil and gas reservoirs

Using supercritical CO2 (sCO2) for geothermal exploitation not only improves the heat extraction rate and saves injection-production energy consumption, but also gains environmental benefit of carbon sequestration. Compared to hot dry rock, depleted oil and gas reservoirs are nature porous reservoirs harboring abundant geothermal resources, in which artificial reservoir fracturing is unnecessary prior to the geothermal extraction. Besides, a lot of pre-existing oil-gas well networks along with ground facilities can be reutilized, significantly reducing geothermal drilling costs. This paper conducts a numerical investigation on using sCO2 for heat extraction in depleted oil and gas reservoirs. Firstly, A coupled wellbore-reservoir model is established to analyze the flow and heat transfer characteristics of sCO2 in the reservoir and injection-production well. Secondly, the variation of sCO2 gas saturation, production temperature, production rate and heat extraction rate are studied. Finally, the influence of different factors on sCO2 heat extraction performance is examined. The results indicate that the maximum sCO2 heat extraction rate is 18.45 MW, and the temperature rise of sCO2 within the reservoir is still higher than 24 °C after 30 years. To enhance sCO2 heat extraction performance, reducing injection temperature, increasing production pressure and well spacing are advisable. In the case of multi-well scheme, a higher ratio of production wells to injection wells and a decentralized arrangement pattern are encouraged. The findings of this paper are anticipated to provide theoretical basis and technical support for efficient geothermal harvest in depleted oil and gas reservoirs and potential appropriate option in building heating.

Fine-Grained Meetup Events Extraction Through Context-Aware Event Argument Positioning and Recognition

Extracting meetup events from social network posts or webpage announcements is the core technology to build event search services on the Web. While event extraction in English achieves good performance in sentence-level evaluation [1], the quality of auto-labeled training data via distant supervision is not good enough for word-level event extraction due to long event titles [2]. Additionally, meetup event titles are more complex and diverse than trigger-word-based event extraction. Therefore, the performance of event title extraction is usually worse than that of traditional named entity recognition (NER). In this paper, we propose a context-aware meetup event extraction (CAMEE) framework that incorporates a sentence-level event argument positioning model to locate event fields (i.e., title, venue, dates, etc.) within a message and then perform word-level event title, venue, and date extraction. Experimental results show that adding sentence-level event argument positioning as a filtering step improves the word-level event field extraction performance from 0.726 to 0.743 macro-F1, outperforming large language models like GPT-4-turbo (with 0.549 F1) and SOTA NER model SoftLexicon (with 0.733 F1). Furthermore, when evaluating the main event extraction task, the proposed model achieves 0.784 macro-F1.

Research on Image Feature Extraction Based on Convolutional Neural Network

Abstract. This article explores the impact of various hyperparameters on the performance of image feature extraction using convolutional neural networks (CNNs), with a focus on learning rate, dropout rate, batch size, and the number of epochs. Using the CIFAR-10 dataset, extensive experiments were conducted to optimize these parameters, aiming to achieve high accuracy while avoiding overfitting. The findings underscore the importance of carefully selecting these hyperparameters to balance training efficiency and model performance. Through a rigorous analysis of the effects of these hyperparameters on model performance under various configurations, including training accuracy, test accuracy, training loss, and test loss, our experimental results indicate that the model achieves optimal performance with a learning rate of 0.0001 and a dropout rate of 0.5. The model demonstrates optimal performance in avoiding overfitting when the number of training epochs is set to 10. Additionally, although batch size has a relatively minor effect on overall model optimization, a slight improvement in performance was observed when the batch size was set to 32.

Effect of impinging angles of liquid–liquid distributor on the extraction and mass transfer performance in a rotating packed bed

Liquid-liquid extraction is a typical separation process that is widely used in various chemical processes. However, the effect of two-liquid feeding mode on the liquid–liquid Murphree efficiency (E) of rotating packed bed (RPB) remains unclear, limiting its applications. In this work, liquid–liquid distributor with different impinging angles (2θ) of two feeding liquid jets in the RPB was designed to evaluate the E, taking water as the extractant to extract benzoic acid from kerosene into the aqueous phase. The effects of operating conditions such as the 2θ, high gravity level (β), and liquid initial velocity (u0) on the E and overall volumetric mass transfer coefficient (KLa) of the RPB were studied. Experimental results revealed that both E and KLa increased with the increase of 2θ, β, and u0. In particular, when 2θ = 90°, E and KLa exhibit optimal values. At β = 49.25, u0 = 1.75 m/s, and 2θ = 90°, since the average diameter of the daughter droplets was smaller and the effective mass transfer area was larger, E reached a maximum of 64.7 % and KLa was 0.489 s−1. A correlation for predicting KLa has been established, with deviations falling within ± 20 %. This work provides theoretical basis for guiding the enhancement of mass transfer in liquid–liquid systems.