A5101960498- Advanced SAR Imaging Techniques
- Non-Invasive Vital Sign Monitoring
- Advanced Chemical Sensor Technologies
- Composite Structure Analysis and Optimization
- Numerical methods in engineering
- Digital Filter Design and Implementation
- Gas Sensing Nanomaterials and Sensors
- Analog and Mixed-Signal Circuit Design
- Image and Signal Denoising Methods
- Advanced Memory and Neural Computing
- Hand Gesture Recognition Systems
- CCD and CMOS Imaging Sensors
- Gait Recognition and Analysis
- Vibration and Dynamic Analysis
- Advanced Numerical Analysis Techniques
- Optical Polarization and Ellipsometry
- Photoreceptor and optogenetics research
- Neuroscience and Neural Engineering
- Analytical Chemistry and Chromatography
- Advanced Adaptive Filtering Techniques
- Radar Systems and Signal Processing
- Numerical Methods and Algorithms
- Advanced Data Compression Techniques
- Analytical Chemistry and Sensors
- Retinal Imaging and Analysis
Shenzhen University
2015-2024
Beijing University of Chinese Medicine
2023-2024
Shanghai University of Engineering Science
2024
Dalian University of Technology
2018-2023
Beijing Institute of Technology
2018-2023
175th Hospital of People's Liberation Army
2017-2019
Xiamen University
2017-2019
South China University of Technology
2019
ShanghaiTech University
2017
Nanfang Hospital
2013-2015
Diabetic retinopathy (DR) is an important cause of blindness worldwide. However, DR hard to be detected in the early stages, and diagnostic procedure can time-consuming even for experienced experts. Therefore, a computer-aided diagnosis method based on deep learning algorithms proposed automatedly diagnose referable diabetic by classifying color retinal fundus photographs into two grades. In this paper, novel convolutional neural network model with Siamese-like architecture trained transfer...
Abstract Neuromorphic computing systems that are capable of parallel information storage and processing with high area energy efficiencies, offer important opportunities for future in‐memory computing. Here, it is shown a carbon dots/silk protein (CDs/silk) blend can be used as light‐tunable charge trapping medium to fabricate an electro‐photoactive transistor synapse. The synaptic device optically operated in volatile or nonvolatile modes, ensuring concomitant short‐term long‐term...
In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in field computer vision, designed with up to 38 layers. general, proposed neural network, named GasNet, consists of: six convolutional blocks, each block consist layers; pooling layer; and fully-connected layer. Together, these various layers make powerful deep model Experimental results show that method is an effective technique classifying...
Many deep learning (DL) models have shown exceptional promise in radar-based human activity recognition (HAR) area. For HAR, the raw data is generally converted into a 2-D spectrogram by using short-time Fourier transform (STFT). All existing DL methods treat as an optical image, and thus corresponding architectures such convolutional neural networks (2D-CNNs) are adopted those methods. These that ignore temporal characteristics ordinarily lead to complex network with huge amount of...
Radar-based human activity recognition (HAR) has great potential in many fields, such as surveillance, smart homes, and human-computer interaction. Complex deep neural networks have brought significant improvement classification performance but also a surge of computational cost the number parameters, which makes it challenging to deploy mobile devices. However, existing studies this area mainly focus on improving accuracy. In article, we propose an extremely efficient convolutional network...
Abstract This study investigates therapeutic efficacy of photothermal therapy (PTT) in an orthotropic xenograft model bone metastasis breast cancer. The near-infrared (NIR) irradiation on Multi-Walled Carbon Nanotubes (MWNTs) resulted a rapid heat generation which increased with the MWNTs concentration up to 100 μg/ml. alone exhibited no toxicity, but inclusion dramatically decreased cell viability when combined laser irradiation. Thermographic observation revealed that treatment 10 μg...
In this paper, we present a novel one-dimensional deep convolutional neural network (1D-DCNN) with multi-label-way-based algorithm for comprehensively and automatically extracting features classifying mixture gases. Although number of pattern recognition methods have been used to analyze the mixed gases, performances these highly depend on hand-crafted feature engineering. By contrast, proposed implementation, based convolution, is capable distinguishing individual component binary gases...
Division of focal plane (DoFP) polarimeter is widely used in polarization imaging sensors. The periodically arranged micro-polarizers integrated on the ensure its outstanding real-time performance, but reduce spatial resolution output images and further affect calculation parameters. In this paper, a four-layer, end-to-end fully convolutional neural network called Fork-Net proposed, which aims to directly improve quality three properties: intensity (i.e., S0), degree linear (DoLP), angle...
To keep pace with the upcoming big-data era, development of a device-level neuromorphic system highly efficient computing paradigms is underway numerous attempts. Synaptic transistors based on an all-solution processing method have received growing interest as building blocks for spikes. Here, we propose and experimentally demonstrated dual operation mode in...
Abstract Parallel information storage coupled with density is a major focus for non‐volatile memory devices to achieve neuromorphic computing that can work at low power. In this regard, photoactive charge‐trapping medium consisting of inorganic heteronanosheets the fabrication synaptic transistor demonstrated. This device senses and responds near‐infrared (NIR) light signals mimics memorization dynamic forgetting process due reversible nature photogenerated charge interaction. Device‐level...
Fast recognition of flammable and toxic gas species within very short response time is a challenging task for the sensing devices adopted in wide range applications. The accuracies previous implementations are always constrained by limited feature or dynamic information extracted from transient curves. In order to address this issue, paper, we propose novel hybrid approach with both convolutional recurrent neural networks combined, which based on long short-term memory module. Featuring...
Previously, the 2-D convolutional neural networks (2-D-CNNs) have been introduced to classify human activity based on micro-Doppler radar. Whereas these methods can achieve high accuracy, their application is limited by computational complexity. In this letter, an end-to-end 1-D network (1-D-CNN) first proposed for radar-based sensors classification. 1-D-CNN, inception densely block (ID-Block) tailored 1-D-CNN proposed. The ID-Block incorporated three techniques: module, dense network, and...
Background: Advanced oxidation protein products (AOPPs), a marker of oxidative stress, are prevalent in many kinds disorders. Rheumatoid arthritis (RA), mainly resulting from the dysfunction fibroblast-like synoviocytes (FLSs), is related to stress. Although increased levels AOPPs RA patients were reported, effect on FLSs function still remains unclear. Therefore, our study aims investigate whether have an inflammatory response vitro. Methods: obtained both knees rats treated with or without...
Human detection and activity classification has recently become a key technology in many applications, e.g., human computer interaction surveillance for public industrial security. In this work, we propose novel end-to-end deep learning-based framework called the Fourier convolutional neural network (F-Convents) to tackle problem. The input of F-ConvNet consists raw frames radar data. It is fed new layer layer, which transforms signal into domain optimized task. A weight initialization...