A5100784394- Electric Motor Design and Analysis
- Magnetic Properties and Applications
- Topic Modeling
- Face recognition and analysis
- Magnetic Bearings and Levitation Dynamics
- Biometric Identification and Security
- Natural Language Processing Techniques
- Digital Media Forensic Detection
- Advanced Steganography and Watermarking Techniques
- Generative Adversarial Networks and Image Synthesis
- Speech Recognition and Synthesis
- Induction Heating and Inverter Technology
- Advanced DC-DC Converters
- Control and Stability of Dynamical Systems
- Arsenic contamination and mitigation
- Speech and Audio Processing
- Advancements in Photolithography Techniques
- Mine drainage and remediation techniques
- Diverse Industrial Engineering Technologies
- Education and Technology Integration
- Advanced SAR Imaging Techniques
- Face and Expression Recognition
- Evacuation and Crowd Dynamics
- Indoor and Outdoor Localization Technologies
- Antibiotic Resistance in Bacteria
Southwest Jiaotong University
2024
National Clinical Research
2024
Zhejiang Business Technology Institute
2024
Hangzhou Dianzi University
2023-2024
Chinese PLA General Hospital
2024
Tianjin University
2024
Tsinghua University
2023
Beijing Institute of Graphic Communication
2023
Wuhan University of Technology
2022
Monolithic Power Systems (United States)
2021
ObjectivesThe objectives of this study were to determine species distribution and in vitro antifungal susceptibility Candida isolates identified the multicentre China-SCAN invasive infection (ICI) intensive care units (ICUs) across China.
Recently, the generative adversarial networks (GANs) generated images have been spread over social networks, which brings new challenge in community of media forensics. Although some reliable forensic tools advanced study detecting GAN images, while detection accuracy cannot be guaranteed when facing malicious post-processing attacks, especially practical network scenario. Thus, this context, we propose a novel well-designed deep neural equipped with handcrafted features for dealing problem....
The film Her features Samantha, a sophisticated AI audio agent who is capable of understanding both linguistic and paralinguistic information in human speech delivering real-time responses that are natural, informative sensitive to emotional subtleties. Moving one step toward more from recent advancement end-to-end (E2E) systems, we propose LUCY, E2E model (1) senses responds user's emotion, (2) deliver succinct natural style, (3) use external tool answer inquiries. Experiment results show...
Underwater images play a critical role in underwater exploration and related tasks. However, due to light attenuation other factors, often suffer from color distortion low contrast, which some extent limit the efficiency safety of exploration. To meticulously address these issues enhance accuracy reliability exploration, this paper proposes multi-task image enhancement method based on Retinex theory. This divides task into several sub-tasks, including decomposition, correction, detail...
This paper evaluates the performance of synchronous reluctance machine (SynRM) and permanent magnet assisted (PMASynRM) for low-cost ventilation application. A unified mathematical model is established with interior (IPMSM) comparison. Besides, permeability frozen method utilized to distinguish basic electromagnetic torque torque. Both efficiency power factor at whole speed region are assessed demonstrate optimal working three analyzed machines.
To evaluate a system for otomicrosurgery based on 4K three-dimensional (3D) exoscope technology and apply it to cochlear implantation.
Due to the rapid development of computing hardware resources and dramatic growth data, pre-trained models in speech recognition, such as Whisper, have significantly improved performance recognition tasks. However, these usually a high computational overhead, making it difficult execute effectively on resource-constrained devices. To speed up inference reduce model size while maintaining performance, we propose novel guided knowledge distillation quantization for large Whisper. The student...
This paper proposes an analytical method to calculate the air-gap flux density and magnetizing inductance of direct quadrature ( d q) axes synchronous reluctance machines (SynRMs). Both inductances are obtained through precise calculation profile, which is achieved by taking advantage conformal mapping Maxwell's equations. A nonlinear magnetic equivalent circuit (MEC) model built take saturation effect into consideration solving iteratively, finally yields potentials on rotor. Such then used...
This paper proposes an analytic equivalent circuit model of the single-secondary trans-inductor voltage regulator (TLVR-1). Such acts as a powerful tool to calculate performance TLVR-1, including phase current ripple, total output slew rate, and maximum flowing through compensation inductor. It is also proved that TLVR-1's conventional multi-phase interleaved buck converter, whose per-phase inductance value much smaller L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML"...
This paper proposes a fast and generalized multiobjective design optimization method for the synchronous reluctance machines (SynRMs). The novel analytical model, based on Maxwell's equations assisted by magnetic equivalent circuit (MEC), is adopted in order to calculate essential performance indices (PIs) including average torque, torque density efficiency. proposed model prevails over prevalent finite element analysis (FEA) terms of calculation speed, while maintains accuracy results....
Target localization and situational awareness of air-ground cooperation have become an important trend inevitable choice in the field unmanned equipment swarm intelligence. The mode framework related algorithms between UAV ground vehicles, personnel, etc. also a research hotspot artificial purpose this study is to explore cheap, novel efficient mode. Specifically, proposes "UAV-Observer Ward" mode, creatively designs new type called "observer ward" (small volume long-distance wireless human...
This paper presents an analytical method to calculate the unsaturated direct (d) and quadrature (q) axis magnetizing inductance of synchronous reluctance machines (SynRMs). The d-axis is calculated by conformal mapping solving partial differential equations magnetic potentials, while q-axis solved from a equivalent circuit (MEC) model. With d q-axes calculated, it will be straightforward obtain electromagnetic torque SynRM. solves much faster than finite element analysis (FEA), yields result...
This study analyzed current college students' learning of English and explicated the advantages applying social media such as Weibo WeChat for learning. The established a bended mode Spiral Sudoku exploiting to allow students learn anywhere anytime. To examine effectiveness model, 34 non-English-major freshmen (NEMF) in ZheJiang University Technology participated one-way quasi-experiment. It was found that all participants loved blended mode, 91.2% them improved on their oral 82.3% reading...
Parameter quantization for Large Language Models (LLMs) has attracted increasing attentions recently in reducing memory costs and improving computational efficiency. Early approaches have been widely adopted. However, the existing methods suffer from poor performance low-bit (such as 2 to 3 bits) scenarios. In this paper, we present a novel effective Column-Level Adaptive weight Quantization (CLAQ) framework by introducing three different types of adaptive strategies LLM quantization....
The cultivation of expertise for large language models (LLMs) to solve tasks specific areas often requires special-purpose tuning with calibrated behaviors on the expected stable outputs. To avoid huge cost brought by manual preparation instruction datasets and training resources up hundreds hours, exploitation open knowledge including a wealth low rank adaptation (LoRA) serves as good starting point. However, existing methods model data selection focus performance general-purpose...
Instruction tuning plays a critical role in aligning large language models (LLMs) with human preference. Despite the vast amount of open instruction datasets, naively training LLM on all existing instructions may not be optimal and practical. To pinpoint most beneficial datapoints, data assessment selection methods have been proposed fields natural processing (NLP) deep learning. However, under context tuning, there still exists gap knowledge what kind evaluation metrics can employed how...
Due to the complex background and less effective information in low-resolution noisy images, detection of small ships SAR images suffers from low rate high false alarm rate. In order solve above problems, this paper proposes a method detect ship targets based on improved YOLOv8. Firstly, deformable convolutional networks are added leading network improve feature extraction. Then, efficient multi-scale attention module is fused with backbone enhance effect targets. The weighted intersection...
In this paper, we propose a robust adaptive Kalman filter design for the dynamic nonlinearity compensation in heterodyne laser interferometer expressed by state equation with parameters uncertainty. The interferometer's measurement process is modeled to improve displacement precision. factor adopted tune covariances of noise and noise. Through comparison experimental results those obtained capacitance sensor, it proved that can robustness system accuracy real time.
The modeling of thermostatically controlled loads (TCLs) is great importance due to the vital role they play in demand response programs. A detailed multiphysics model an air-conditioned house proposed this paper. physically based approach used for constructing model, which includes both thermal and air conditioner (AC) electrical modeling. mathematical expresses laws governing operation TCL. An example system simulated evaluate performance with realistic AC parameters. accurately captures...
Face recognition is a widely-used technique for identification or verification, where verifier checks whether face image matches anyone stored in database. However, scenarios the database held by third party, such as cloud server, both parties are concerned about data privacy. To address this concern, we propose CryptoMask, privacy-preserving system that employs homomorphic encryption (HE) and secure multi-party computation (MPC). We design new encoding strategy leverages HE properties to...