A5100399361- EEG and Brain-Computer Interfaces
- Blind Source Separation Techniques
- High voltage insulation and dielectric phenomena
- Neuroscience and Neural Engineering
- Functional Brain Connectivity Studies
- Sparse and Compressive Sensing Techniques
- Adversarial Robustness in Machine Learning
- Emotion and Mood Recognition
- Thermal Analysis in Power Transmission
- ECG Monitoring and Analysis
- Advanced Neural Network Applications
- Water Systems and Optimization
- Radiation Effects in Electronics
- Gaze Tracking and Assistive Technology
- Neural dynamics and brain function
- Advanced Adaptive Filtering Techniques
- Cavitation Phenomena in Pumps
- Advanced Memory and Neural Computing
- Icing and De-icing Technologies
- Impact of Light on Environment and Health
- Optical Wireless Communication Technologies
- Microwave Imaging and Scattering Analysis
- Antimicrobial agents and applications
- Hydraulic flow and structures
- Spectroscopy and Chemometric Analyses
Tianjin University
2016-2023
ShanghaiTech University
2022
South China Normal University
2020-2022
Xidian University
2021
Zero to Three
2020
Beijing Normal University
2020
Akita Prefectural University
2019
University of Jinan
2017
Northwest Institute of Nuclear Technology
2016
Xi'an Jiaotong University
2016
Objective. Modern motor imagery (MI)-based brain computer interface systems often entail a large number of electroencephalogram (EEG) recording channels. However, irrelevant or highly correlated channels would diminish the discriminatory ability, thus reducing control capability external devices. How to optimally select and extract associated features remains big challenge. This study aims propose validate deep learning-based approach automatically recognize two different MI states by...
Classification of electroencephalogram-based motor imagery (MI-EEG) tasks raises a big challenge in the design and development brain-computer interfaces (BCIs). In view characteristics nonstationarity, time-variability, individual diversity EEG signals, deep learning framework termed SSD-SE-convolutional neural network (CNN) is proposed for MI-EEG classification. The consists three parts: 1) sparse spectrotemporal decomposition (SSD) algorithm feature extraction, overcoming drawbacks...
Machine fault diagnosis collects massive amounts of vibration data about complex mechanical systems. Performing feature detection from these sets has already led to a major challenge. Compressive sensing theory is new sampling framework that provides an alternative the well-known Shannon theory. This enables recovery sparse or compressible signals small set nonadaptive linear measurements. However, it suboptimal recover whole compressive measurements and then solve identification problems...
Classification of electroencephalogram-based motor imagery (MI-EEG) tasks is crucial in brain computer interfaces (BCI). In view the characteristics non-stationarity, time-variability and individual diversity EEG signals, a novel framework based on graph neural network proposed for MI-EEG classification. First, an adaptive convolutional layer (AGCL) constructed, by which electrode channel information are integrated dynamically. We further propose spatiotemporal (ASTGCN), fully exploits...
Convolutional neural networks (CNNs) are widely used in computer vision and natural language processing. Field-programmable gate arrays (FPGAs) popular accelerators for CNNs. However, if critical applications, the reliability of FPGA-based CNNs becomes a priority because FPGAs prone to suffer soft errors. Traditional protection schemes, such as triple modular redundancy (TMR), introduce large overhead, which is not acceptable resource-limited platforms. This article proposes use an ensemble...
In the human-computer interaction (HCI), electroencephalogram (EEG) access for automatic emotion recognition is an effective way robot brains to perceive human behavior. order improve accuracy of recognition, a method EEG based on deep hybrid network was proposed in this paper. Firstly, collected decomposed into four frequency band signals, and multiscale sample entropy (MSE) features each were extracted. Secondly, constructed 3D MSE feature matrices fed autonomous learning. The composed...
Thermal aging is a common form of cable deterioration. In this paper, the effect thermal on cables evaluated by analyzing harmonic characteristics in leakage currents. Cable samples were first fabricated and subjected to accelerated tests at 120 °C. The experimental circuits built test dielectric loss factor AC current different times. Then, improved variational modal decomposition (VMD) algorithm was used for time–frequency analysis current, relationship between harmonics investigated....
The vertical pipe inlet/outlet, with a horizontal plate widely used in the pumped storage plant, has characteristics of complex two-direction flow. experiment was conducted on original shape at different discharges. Compared experimental results, current study presents three-dimension numerical investigation turbulent flow inlet/outlet using realizable k-ϵ model. hydraulic such as velocity distribution and head losses diversion orifices heights divergence angles have been analysed. results...
Autism spectrum disorder (ASD) is a pervasive neurodevelopmental characterized by restricted interests and repetitive behaviors. Non-invasive measurements of brain activity with functional magnetic resonance imaging (fMRI) have demonstrated that the abnormality in default mode network (DMN) crucial neural basis ASD, but time-frequency feature DMN has not yet been revealed. Hilbert-Huang transform (HHT) conducive to extraction biomedical signals recently suggested as an effective way explore...
Epilepsy is a neurological disease caused by ab-normal neural electrical discharges. Electroencephalography (EEG) powerful tool to measure the brain activity and has been widely used for seizure detection. Manual EEG analysis labor-intensive time-consuming. Automatic detection urgently demanded long-time monitoring. Many methods have proposed automatic based on signals. However, most of existing are patient-specific with limited generaliz-ability. Few studies investigate inter-patient...
Convolutional Neural Networks (CNNs) are widely used in image classification tasks. To fit the application of CNNs on resource-limited embedded systems, pruning is a popular technique to reduce complexity network. In this paper, robustness pruned network against errors parameters examined with VGG16 as case study. The effects weights, bias, and batch normalization (BN) evaluated for different rates based error injection experiments. results show that general networks more weights robust...
The lateral inlet/outlet plays a critical role in the connecting tunnels of water delivery system pumped hydroelectric storage (PHES). Therefore, shape was improved through computational fluid dynamics (CFD) optimization based on optimal surrogate models. CFD method applied this paper validated by physical experiment that carefully designed to meet bidirectional flow requirements. To determine good compromise between generation and pump mode, reasonable weights were defined better evaluate...
Convolutional Neural Networks (CNNs) are widely used in computer vision and natural language processing. Due to large computational requirements, implementation of CNNs on FPGAs becomes an popular option. As being safety critical applications, reliability become a priority. This poses challenges as prone suffer soft errors. Traditional fault tolerant techniques based modular redundancy introduce overhead, which may not be acceptable for many resources-limited embedded system. paper explores...
The leakage current (LC) is directly related to the discharge process on outdoor insulator surface, which considered as one of critical measurements for flashover prediction ice-covered insulators. This paper conducts icing experiments in an artificial cold room and uses a high-speed camera photograph dynamic arcs. Image processing techniques were applied extract arc perimeter. characteristics LC perimeter analyzed at each stage DC process, well relationship between during development....
Objective.Energy consumption is a critical issue in resource-constrained wireless neural recording applications with limited data bandwidth. Compressed sensing (CS) has emerged as powerful framework addressing this owing to its highly efficient compression procedure. In paper, CS-based approach termed simultaneous analysis non-convex optimization (SANCO) proposed for large-scale, multi-channel local field potentials (LFPs) recording.Approach.The SANCO method consists of three parts: (1) the...
In generative modeling, tokenization simplifies complex data into compact, structured representations, creating a more efficient, learnable space. For high-dimensional visual data, it reduces redundancy and emphasizes key features for high-quality generation. Current methods rely on traditional autoencoder framework, where the encoder compresses latent decoder reconstructs original input. this work, we offer new perspective by proposing denoising as decoding, shifting from single-step...
Motivation: Understanding infant neurodevelopment is pivotal for unraveling the anatomical underpinnings of psychomotor and cognitive functions, as well pinpointing origins various disorders. Goal(s): Introduce an integrated multi-modality MRI data processing pipeline tailored development studies, with goal reliablly discerning relationship across brain anatomy functions. Approach: Incorporating precise deep learning tools specifically designed brain, structural, functional, diffusion can be...
Convolutional Neural Networks (CNNs) have been widely used for image recognition or natural language processing. When CNNs are in safety-critical applications, their reliability becomes a priority and particular tolerance to soft errors. Unfortunately, traditional fault tolerant techniques based on modular redundancy introduce large overhead, which is not acceptable many resource-limited embedded systems given the complexity of CNNs. To reduce cost protecting CNNs, use an ensemble smaller...
In order to improve the stability and reliability of transmission lines, this paper investigates growth process surface discharge characteristics icicles on outdoor insulator strings under DC voltages. a charged icing experiment in an artificial cold room, maximum icicle lengths at different shed edges were recorded measured by using high-speed camera, influence discharges was investigated. Under voltages, finally attain stable length. However, rate is faster positive voltage than that...
In this paper, a method for classifying electroencephalographic (EEG) recordings with images as stimulation is introduced, which aims at selecting the target images.EEG to be processed are referred onset of test single so avoid spending extra time on repeating images.Independent component analysis (ICA) used reduce redundancy EEG recordings, and wavelet packet (WP) efficient dealing non-stationary character brain activity.Feature vectors extracted by that combines these two algorithms.The...
Task-based fMRI has been widely used to guide presurgical mapping non-invasively. Application of resting-state (rs-fMRI) was recently reported in as well (Zhang et al., 2009) patient severe clinical conditions may not be able perform tasks. Independent component analysis (ICA)-based (Kokkonen shows its advantage no a priori information is required, which naturally fits for rsfMRI. However, such issues should neglected that components identified by subjective visual inspection often leads...