A5102966656- Radar Systems and Signal Processing
- Direction-of-Arrival Estimation Techniques
- Advanced SAR Imaging Techniques
- Microwave Imaging and Scattering Analysis
- Speech and Audio Processing
- Indoor and Outdoor Localization Technologies
- Geophysical Methods and Applications
- Antenna Design and Optimization
- Blind Source Separation Techniques
- Underwater Acoustics Research
- Magnetic Bearings and Levitation Dynamics
- Sparse and Compressive Sensing Techniques
- Radio Wave Propagation Studies
- Advanced Adaptive Filtering Techniques
- Electric Motor Design and Analysis
- Structural Health Monitoring Techniques
- Advanced Wireless Communication Technologies
- Model Reduction and Neural Networks
- Hearing Loss and Rehabilitation
- Distributed Sensor Networks and Detection Algorithms
- Stochastic processes and financial applications
- Gaussian Processes and Bayesian Inference
- Radiomics and Machine Learning in Medical Imaging
- Non-Invasive Vital Sign Monitoring
- Ultrasonics and Acoustic Wave Propagation
Ningbo No. 2 Hospital
2025
Xidian University
2015-2024
First Affiliated Hospital of Zhengzhou University
2024
Midea Group (China)
2023-2024
Oak Ridge National Laboratory
2023-2024
Xi'an Jiaotong University
2023-2024
Tongji University
2022-2023
Chinese Academy of Sciences
2023
Jiangnan University
2022
Sun Yat-sen University
2022
A new Unitary ESPRIT algorithm for joint direction of departure (DOD) and arrival (DOA) estimation in bistatic MIMO radar is proposed. The properties centro-Hermitian matrices are utilised to transform the complex-valued data matrix into a real-valued matrix. Then rotational invariance equations signal subsapce figured out estimate DOAs DODs which paired automatically way. proposed provides increased accuracy with reduced computational complexity owing processing double number samples...
A new linear sparse array based on the nested is proposed, which enjoys all good properties of two-level array, and can provide more degrees freedom (DOF). The constructed by two uniform arrays (ULAs) an additional sensor. sensor locations, aperture, achievable DOF from its difference co-array (DCA) are benefited for closed-form expressions. Furthermore, resulting DCA kept as a hole-free ULA. optimal numbers sensors in ULAs provided total number physical derived. This resolve sources achieve...
A reverberation-time-aware deep-neural-network (DNN)-based speech dereverberation framework is proposed to handle a wide range of reverberation times. There are three key steps in designing robust system. First, contrast sigmoid activation and min-max normalization state-of-the-art algorithms, linear function at the output layer global mean-variance target features adopted learn complicated nonlinear mapping from reverberant anechoic improve restoration low-frequency intermediate-frequency...
We propose an integrated end-to-end automatic speech recognition (ASR) paradigm by joint learning of the front-end signal processing and back-end acoustic modeling. believe that "only good can lead to top ASR performance" in challenging environments. This notion leads a unified deep neural network (DNN) framework for distant achieve both high-quality enhanced high-accuracy simultaneously. Our goal is accomplished two techniques, namely: (i) reverberation-time-aware DNN based dereverberation...
We propose anew antenna array design approach for a multiple-input and multiple-output (MIMO) radar, which has closed-form expressions the sensor locations number of achievable degrees freedom (DOFs). This new utilizes nested as transmitting receiving arrays. employ difference coarray sum (DCSC) MIMO radar to obtain more DOFs direction-of-arrival (DOA) estimation. Via properly designing interelement spacings arrays, we can hole-free DCSC. The characteristics geometries are analyzed optimal...
Occult nodal metastasis (ONM) plays a significant role in comprehensive treatments of non-small cell lung cancer (NSCLC). This study aims to develop deep learning signature based on positron emission tomography/computed tomography predict ONM clinical stage N0 NSCLC. An internal cohort (n = 1911) is included construct the (DLNMS). Subsequently, an external 355) and prospective 999) are utilized fully validate predictive performances DLNMS. Here, we show areas under receiver operating...
We present a supervised learning framework of training generative models for density estimation. Generative models, including adversarial networks (GANs), normalizing flows, and variational auto-encoders (VAEs), are usually considered as unsupervised because labeled data unavailable training. Despite the success there several issues with training, e.g., requirement reversible architectures, vanishing gradients, instability. To enable in we utilize score-based diffusion model to generate...
To mitigate the overlapping of radar and communication frequency bands caused by large-scale devices access, we propose a novel integrated sensing (ISAC) system, where micro base station (MiBS) simultaneously carries out both target cooperative communication. Concretely, MiBS, acting as equipment, can also serve full-duplex decode-and-forward relay to assist end-to-end Moreover, nonorthogonal downlink transmission (NO-DLT) is adopted between macro Internet Things devices, so that spectrum...
In a low-angle target parameter estimation scenario, the backscattered signals from targets are often distorted due to clutter and multipath, which significantly degrades performance of direction-of-arrival (DOA) estimation. general, multipaths modeled as coherent with respect direct path. Decorrelation algorithms such spatial smoothing matrix reconstruction can mitigate effect multipath. However, most these methods will perform poorly or even fail for solving problem in complex terrain...
To improve precision in source localization from a time difference of arrival (TDOA) that has large measurement errors, this paper proposes TDOA positioning algorithm based on an improved two-step constrained total least-squares algorithm; the comprise iterative technique alternating direction method multipliers (ADMM). The linearizes obtained observation equation, and by considering influence errors its accuracy, it provides optimal model. It first transforms initial unary optimization...
In this paper, we propose a unified array geometry, dubbed generalized nested subarray (GNSA), for the underdetermined direction-of-arrival estimation. The GNSA is composed of multiple, identical subarrays, which can be minimum redundancy (MRA), (super) array, uniform linear (ULA), or any other arrays with hole-free difference coarrays (DCAs). By properly design spacings between resulting DCA also (filled) ULA. When an MRA and meanwhile its sensors' positions follow configuration, (NMRA)...
A novel direction of arrival (DOA) estimation method is proposed for very high‐frequency (VHF) radar by the deep neural network (DNN) under strong multipath effect and complex terrain environment. The classical methods are all based on signal model, hence, it often causes problem model mismatch results in poor performance estimation. It generally considered that serious reduces precision elevation However, characteristics exploited used to improve this study. This highlight method. approach...
In this paper, a multiscale sparse array, which is composed of spatially-spread electromagnetic-vector-sensors (SS-EMVSs), proposed to estimate the direction-of-arrivals (DOA) and polarizations multiple sources. The SS-EMVS three orthogonally oriented but spatially noncollocated dipoles measure electric field loops magnetic field, simultaneously. an array SS-EMVSs placed along y-axis, two sub-arrays, i.e., first n <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...
In low-altitude target situation, the multi-path signals cause amplitude-phase distortion of direct signal from targets and degrade performance existing methods. Hence, in this paper, we propose a phase enhancement method for low-angle estimation using supervised deep neural network (DNN) to mitigate distortion, thus improve direction arrival (DOA) accuracy. The mapping relationship between original difference distribution received desired is learned by DNN during training. test data...
Time reversal can resolve multipath problems and increase the detection probability, but performance decreases when target is moving. Because of change in channel response due to motion, focusing effect a time-reversal-transmitted signal reduced. In this article, novel method based on time reversal, which detects moving embedded environment with variation, proposed. We establish conventional time-reversal models an indoor diffuse derive likelihood ratio tests under these two conditions. The...
Fall recognition is very important for the elderly. Consequently, fall using convolution neural networks has been widely studied. However, current studies rarely consider network implementation on radar devices. As devices have limited processing power and storage space, with high computational complexity, time-consuming, parametric quantities will limit their terminal applications. Hence, requires a lightweight convolutional (LWCNN). With regard to this, an LWCNN comprised of channel...