Search

Neuroimaging technique is a powerful tool to characterize the abnormality of brain networks in schizophrenia. However, the neurophysiological substrate of schizophrenia is still unclear. Here we investigated the patterns of brain functional and structural changes in female patients with schizophrenia using elastic net logistic regression analysis of resting-state functional magnetic resonance imaging data. Data from 52 participants (25 female schizophrenia patients and 27 healthy controls) were obtained. Using an elastic net penalty, the brain regions most relevant to schizophrenia pathology were defined in the models using the amplitude of low-frequency fluctuations (ALFF) and gray matter, respectively. The receiver operating characteristic analysis showed reliable classification accuracy with 85.7% in ALFF analysis, and 77.1% in gray matter analysis. Notably, our results showed eight common regions between the ALFF and gray matter analyses, including the Frontal-Inf-Orb-R, Rolandic-Oper-R, Olfactory-R, Angular-L, Precuneus-L, Precuenus-R, Heschl-L, and Temporal-Pole-Mid-R. In addition, the severity of symptoms was found positively associated with the ALFF within the Rolandic-Oper-R and Frontal-Inf-Orb-R. Our findings indicated that elastic net logistic regression could be a useful tool to identify the characteristics of schizophrenia -related brain deterioration, which provides novel insights into schizophrenia diagnosis and prediction.

An 18–56-GHz gallium nitride (GaN) low-noise amplifier (LNA) monolithic microwave integrated circuit (MMIC) is presented in this letter. This LNA is fabricated with commercial 0.1- $\mu \text{m}$ gate-length GaN-on-silicon (GaN/Si) process. The gain is 16–21.5 dB and the $\vert {S} 11\vert $ and $\vert {S} 22\vert $ are below −5 dB across the band. The noise figure (NF) is 2.2–4.4 dB and the output power at 1-dB compression point ( ${P} _{\text {1 dB}}$ ) is 10–20 dBm. This LNA has four-stage-cascade and common-source structure with $4.8\times 1$ mm2 chip area. The dc power dissipation is 1.4 W. To the best of the author’s knowledge, there has been no GaN LNA covering such a wide bandwidth at this frequency range. This LNA shows ultrawide bandwidth, low NF, high linearity, and high robustness. Moreover, it can be directly integrated with GaN power amplifiers (PAs) and switches to realize the monolithic RF front-ends, which can largely reduce the packaging loss and the module volume. This LNA has a great potential in wideband, high-speed, high-reliability, and high-integration applications.

Benefited from the high breakdown voltage and low noise characteristics, GaN high-electron mobility transistors (HEMTs) can be used for manufacturing of robust low noise amplifiers (LNAs). Therefore, limiter circuits which protect the entire system are no longer necessary and systems with smaller volume can be realized. In this letter, a LNA with low noise figure (NF) and longtime survivability for high input power stress is designed and fabricated in a 100-nm GaN process. The LNA achieves a minimum NF of 0.8 dB and a flat gain of 21 ± 0.5 dB from 18 to 31 GHz. The LNA can survive in a 28-30-dBm input stress for 60 min without significant performance degradation.

Objectives This paper is aimed to explore the value of double source CT angiography (DS-CTA) for diagnosing in-stent restenosis in lower limb artery. Methods From January 2016 to October 2018, all patients with stent in lower limb artery in our hospital were investigated by both DS-CTA and digital subtraction angiography. We measured the minimum lumen diameter and the diameter of the proximal normal vessels under each stent placement. The in-stent restenosis is defined as restenosis when the lumen area decreased by more than 50%. Digital subtraction angiography was performed within 1 week after DS-CT scan. Relationship between DS-CTA and digital subtraction angiography for diagnosing in-stent restenosis in lower limb artery was analyzed. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DS-CTA for diagnosis of in-stent restenosis were analyzed with digital subtraction angiography as the reference standard. A total of 68 stents were placed in 51 patients. Among these patients, 27 cases were diagnosed as in-stent restenosis, presenting as endovascular contrast agent bias or crescent filling defect with the lumen area reducing over 50%, 6 cases of which had no significant in-stent restenosis by digital subtraction angiography analysis. Furthermore, 12 cases were occlusion, in which there was no high density contrast agent in stents; the remaining 41 stents were unobstructed and the contrast agent was filled well, 8 cases of which had significant in-stent restenosis by digital subtraction angiography analysis. In addition, four stents were deformed or distorted. Statistical analysis demonstrated the concentrations of DS-CTA and digital subtraction angiography in diagnosing in-stent restenosis for lower limb artery were closely related, and the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DS-CTA were 72.4%, 84.6%, 77.8%, 80.5%, and 79.4%, respectively. Conclusion DS-CTA has a potential reliability for diagnosis of in-stent restenosis in lower limb artery, which may be further improved to be used for clinical interventional treatment of vascular diseases.

A 23–30-GHz gallium nitride (GaN) low-noise amplifier (LNA) with a noise figure (NF) of 0.87–1.51 dB is presented in this article. This LNA was fabricated with 100-nm gate-length GaN-on-silicon (GaN/Si) microwave monolithic integrated circuit (MMIC) process. The linear gain is 14–17 dB within the band. For investigating the robustness of this LNA, 1-W continuous wave (CW) at 27 GHz was stressed on the input port of the LNA. The gain decreased, and the NF increased after stress. Experimental research and first-principles calculations were carried out to investigate the physical origin of the degradation. The dehydrogenation of $\text{V}_{{\text {Ga}}}$ -H3 complexes causes the decrease of gain, and the creation of $\text{V}_{{\text {Al}}}$ -H4 in the AlN barrier is supposed to cause an increase of NF.

Background DDMS is a rare disease diagnosed by clinical and radiological characteristics. But the complexity of radiological and clinical manifestations of DDMS has become a challenge diagnostically. To date, the reported cases with DDMS had highly varied clinical manifestations including seizures, contralateral hemiplegia/hemiparesis, facial asymmetry, mental retardation, etc. In addition to typical clinical findings, some new characteristics have been recently added to the spectrum of DDMS. However, few cases have been reported to be associated with neuropsychiatric symptoms according to the literature. This study aimed to investigate the neuropsychiatric manifestations associated with Dyke-Davidoff-Masson syndrome (DDMS) and related imaging findings. Methods This study included 7 patients diagnosed with DDMS between 2014 and 2020. The clinical characteristics, neuropsychiatric manifestations, and radiological results were retrospectively evaluated. Results Seven patients (five males and two females) with a mean age of 28.0 ± 9.73 (range 15.0–41.0) years were included. Five patients were admitted to the psychiatric unit due to psychological and behavioral disorders. Two patients were referred to the neurology unit mainly due to epilepsy. Six patients had epileptic seizures, 4 had hemiplegia, 3 had mental retardation, 2 patients had external ear deformities, and 2 had facial asymmetry. Neuropsychiatric symptoms were presented in 6 (85.7 %) cases. Cases 2–6 developed affective disorders. Deficits in verbal communication, impairment of social interaction, lack of insight, adulia and hypobulia appeared in cases 1–4. Schizophrenia with apathy, and epileptic schizoid psychosis were observed in cases 4 and 5 respectively. Case 6 had behavioral disorders, hyperactivity, tic disorder, mental retardation, anxiety, catatonic symptoms and suicidal tendency. Case 7 had seizures and mental retardation, and no psychiatric symptoms were presented. Radiological examinations showed unilateral cerebral atrophy, enlarged lateral ventricles, and various compensatory hypertrophy of the skull in all cases. The midline structure has shifted to the affected side in 5(71.4 %) cases. Atrophy of the basal ganglia or brain stem was observed in 4(57.1 %) cases. Conclusions The hallmark imaging manifestations of DDMS facilitated the diagnosis in most cases. This study illustrated that a variety of psychoneurotic disorders and ear abnormalities were correlated with DDMS.

Considering the uncertainty and multi-disturbance in wind energy conversion system (WECS) and the randomness and instability of the wind power, a full-order sliding mode control (FOSMC) strategy based on an extended state observer (ESO) is proposed for the permanent magnet synchronous generator (PMSG)-based WECSs. First, based on the active disturbance rejection control (ADRC) theory, an ESO is developed to estimate the unknown states, uncertainties in the internal machine parameters and external torque disturbance. Second, a full-order sliding mode controller is designed to ensure the convergence of all system states with much reduced chattering phenomenon. Finally, the effectiveness of the proposed ESO-based FOSMC is confirmed by simulation results. The proposed FOSMC strategy will improve the efficiency and the reliability of the WECS.

Optical Braille Recognition methods usually use many designed steps, such as image deskewing, Braille dots detection, Braille cell grids construction and Braille character recognition, which are less robust for complex Braille scenes. This paper proposes an optimal semantic segmentation framework BraUNet to directly detect and recognize Braille characters in the whole original Braille images. BraUNet adds extra auxiliary learning strategy to UNet network, which uses long-range connections of feature maps between encoder and decoder to get more low-level features. And auxiliary learning strategy can combine multi-class Braille characters segmentation with Braille foreground extraction, which can improve the feature learning ability and the Braille segmentation performance. Then morphological post-processing is used on semantic segmentation results to get the final individual Braille character regions. Experimental results show the proposed framework is robust, effective and fast for Braille characters segmentation and recognition on both complex double sided Braille image dataset and handwritten Braille image dataset.

Identity authentication is of great importance in the digital age where ID information is commonly used in finance, insurance, transportation and other fields. Challenges of identity authentication lie in the verification of the ID card provided by the user and the information extraction from the user’s ID card. To meet the challenge, an identity authentication framework is proposed, which can extract and verify personal information through face verification and ID image recognition. The identity authentication is realized by the proposed face verification model which is called Inception-ResNet Face Embedding (IRFE). IRFE uses an Inception-ResNet structure to ensure a good feature extraction aiming at accurate face verification. Moreover, a robust ID card extraction method named Morphology Transformed Feature Mapping (MTFM) is proposed to extract ID information. Experimental results demonstrate that the proposed IRFE and MTFM outperform state-of-the-art methods both in face verification and in ID extraction.

A 22–30-GHz gallium nitride (GaN) low-noise amplifier (LNA) having a noise figure (NF) of 0.4–1.1 dB is presented in this letter. This LNA is fabricated with 0.1 gate-length GaN-on-silicon high-electron mobility transistor process, which is believed to have the lowest NF among GaN LNA at this frequency range reported to date. The small-signal gain is between 19.5 and 22.5 dB across the band. The output-referred 1-dB compression point (P1dB) and output-referred third-order intercept point (OIP3) are at 20.8- and 34.5-dBm level. The high robust was proven by stressing the LNA with a continuous wave input power of 30 dBm for 1 min. The three-stage LNA is $1.7\times1.3$ mm2 in area and consumes 210-mW dc power. Compared with the traditional gallium arsenide and indium phosphide LNA, the GaN monolithic microwave integrated circuit LNA in this letter exhibits a competitive NF but has much higher robust and linearity.

A 23-31 GHz Monolithic Microwave Integrated Circuit (MMIC) low-noise Amplifier (LNA) based on AlGaN/GaN on Si technology from OMMIC is presented in this paper. Common-source topology with inductive source feedback is utilized for simultaneous noise and input match. Measurement results show that the LNA has a gain of more than 22 dB while achieving an average noise Figure (NF) of 1.1 dB over the designed band and the minimum value of 0.93 dB at 27 GHz. The three stage topology achieves high linearity, providing the 1-dB compression point output power (P}$_{1dB}$) of 23 dBm and a Saturated Output Power (Psat) of $\sim28$ dBm. The robustness of this LNA was tested with 1 Watt continuous-wave input power at 27 GHz. The LNA survived after stress without obvious degradation. Compared with the traditional GaAs LNA, the GaN LNA has competitive NF and much higher linearity.

A 22-30 GHz gallium nitride (GaN) low-noise amplifier (LNA) with a noise Figure (NF) of 0.87-1.51 dB is presented in this work. This LNA was fabricated with 100 nm gate-length AlN/GaN on silicon (Si) microwave monolithic integrated circuit (MMIC) process. The linear gain is 14-17 dB with input/output return loss over 10 dB across the band. To investigate the robustness of this LNA, 1 Watt continuous wave (CW) at 27 GHz was stressed on the input port of LNA. The decrease of gain and increase of NF were found after stress. Experimental research and first-principles calculations were carried out to investigate the physical mechanism of theses degradations, which indicate that the dehydrogenation of VGaH3 complexes in GaN channel caused the decrease of gain and the creation of VAl-H4 in the aluminum nitride (AlN) barrier caused the increase of NF.

In this report, a broad band (15 GHz to 34 GHz) low noise amplifier using 100 nm GaN on silicon technology is designed and fabricated. The LNA shows an extremely low noise figure of 1.2 dB and small signal gain of 18.5±1.5 dB across the bandwidth. The LNA can work in the drain bias from 3.5 V to 8 V without increasing the noise figure. The chip size is 2 mm × 1.3 mm. The robustness of the LNA is also tested by stressing the working LNA with a continuous-wave input power, no significant degradation is observed.

As quadruped robots are suitable to be used in unstructured environment, environment perception for them becomes a popular field of research nowadays. This paper presents a novel and effective detection and tracking system based on stereo camera for quadruped robots, which allows a particular user to designate himself as the robot's leader. Firstly, the object detection is carried out by YOLO-V3 method based on the newfangled residual network. Then, the object tracking is performed based on ECO method, which is an improved kernel algorithm. Finally, a IDLA and CCF cascaded pedestrian re-identification method is adopted to identify the navigator among the crowd. In this work, we tackle the key causes behind the problems of computational complexity and over-fitting, with the aim of simultaneously improving both speed and performance. The comprehensive experimental results validate the effectiveness and robustness of our algorithm. In addition, our system runs at a real-time of 30 frames per second.

We report 100-nm gate AlGaN/GaN high electron mobility transistors (HEMTs) without electron-beam (e-beam) lithography featuring current gain cut-off frequency ( $f_{\mathrm{T}}$ ) and maximum oscillation frequency ( $f_{\max}$ ) of 66 GHz and 206 GHz, respectively. Fabricated in commercial fab, the gate width of ∼80 nm using i-line stepper and sidewall technology on 6-inch wafer indicates the possibility application in rapid mass production compared to slow and expensive e-beam method.

A 24.5-27 GHz gallium nitride (GaN) power amplifier (PA) is reported in this work. This PA was designed with $0.1\ \ \mu \mathrm{m}$ gate-length GaN-on-silicon (GaN/Si) microwave monolithic integrated circuit (MMIC) process. This PA has a 22-25 dB linear gain with high input/output return loss. The measured saturation output power is 35-36 dBm in pulsed work condition (10% duty cycle with 1 ms period). The total drain current is ∼827mA in saturation working condition with drain efficiency of 33%-40%. The area of this PA is $3.9\times 1.9\ \text{mm}^{2}$ with substrate thickness of $100\ \mu \mathrm{m}$ . Compared with the traditional GaN-on-silicon carbon (GaN/SiC) PA, The PA in this work shows comparable output power, high efficiency, but lower cost. This PA has a great advantage in wide commercial market, such as the next generation 5G communication and other high-speed data transmission networks.

A low noise amplifier (LNA) microwave monolithic integrated circuit (MMIC) and a power amplifier (PA) MMIC in Ka band based on 100 nm gallium-nitride on silicon (GaN/Si) high electron mobility transistor (HEMT) process are reported in this work. The LNA has linear gain of 14–17 dB, noise figure (NF) of 0.8–1.1 dB and 1-dB compression output power (P1dB) of 21 dBm in 23–30 GHz frequency band. This is the first public report of GaN LNA having NF below 1 dB at this frequency range. This paper also reported a PA working in 23.5–30 GHz with linear gain of 20 dB, saturated output power of 7.6–12.4 Watt and power added efficiency (PAE) of 24%–37%. The traditional PA is based on gallium-nitride on silicon-carbon (GaN/SiC) process, there has been no PA as high as 10 W level based on GaN/Si process reported in the 20–30 GHz frequency region.

Typical hidden Markov model (HMM) is very sensitive to initial parameters while using random parameters to train HMM often falls into local optimum, and the effect is poor when applied to fault diagnosis. Fault diagnosis algorithm based on simulated annealing algorithm (SA) and hidden Markov model is proposed in this paper. Selecting the best simulated annealing algorithm parameters and combining with Baum-Welch algorithm, we optimize the hidden Markov model and apply it to the rolling bearing fault diagnosis. Experimental results show that the new algorithm for fault diagnosis accurate rate has increased significantly.

Machine Learning Computers (MLCs) with tensor functional units (e.g., NVIDIA's Tensor Core, Google's TPU and Habana's Tensor Processor Core) have emerged significantly over recent years. The broad diversity of MLCs makes it hard to deploy machine learning workloads with optimized performance. Though deep learning compilers (e.g., TVM) are effective to produce optimized code for different hardware back-ends, when deploying to a new MLC, it is tedious to implement platform-specific compilation optimizations by thoroughly understanding system/architectural details. To address this problem, we propose a holistic approach to achieve one-size-fits-all compilation optimization across different MLCs or inference. The key observation is that diverse MLCs share multiple key architectural characteristics for tensor processing, which can be generalized for conducting cross-platform compilation optimizations. Concretely, we propose the Tensor Abstract Machine (TAM), which features such common architectural characteristics, as the abstraction of a broad range of MLCs. To leverage architectural characteristics of the TAM, we propose the Tensor Scheduling Language (TSL) consisting of tensor computation description and tensor scheduling primitives for implementing operations with portable optimization. Experimental results demonstrate that the code generated from the same optimization schedule achieves 1.05x to 2.05x better performance than hand-tuned libraries and deep learning compilers across different platforms.

The early stage degradation of electrical properties in AlGaN/GaN high electron mobility transistors (HEMTs) under fast neutron irradiation is studied. After the 1 MeV neutron irradiation at a low fluence of 1 × 1014 neutrons/cm2, the reverse leakage current decreases while the output and transfer characteristics remain unchanged, which cannot be explained by the previously reported high-fluence degradation model. By employing temperature-dependent gate leakage current measurements, we show that the dislocation related Poole–Frenkel (PF) emission dominates the gate leakage mechanism before and after irradiation whereas the barrier height for electron emission to conductive dislocation increases after the neutron irradiation. A model with the evolution of dislocation from the VGa-decorated configuration to the pure configuration is proposed to describe the degradation of AlGaN/GaN HEMTs at the low-fluence early stage neutron irradiation. This model enriches the understanding of the degradation mechanism of neutron irradiated AlGaN/GaN HEMTs.

This paper reports on an in-depth research on the etching-process induced current degradation in AlGaN/GaN heterostructures by experimental research and theoretical calculations. The channel current degrades by 60% after the etching process and recovers to 90% after thermal annealing at 400 °C. It is found that fluorine is incorporated into the heterostructures of both AlGaN and GaN layers during the etching process. The degradation and recovery of the current is accompanied by the change in the concentration of fluorine in both AlGaN and GaN layers during the experiment. First-principles defect calculations of fluorine in GaN indicate that the dominant defect configuration of F is negatively charged interstitial (Fi−) with the lowest formation energy. The accompanied technology computer-aided design simulation of the energy bands reveals that Fi− raises the conduction band minimum of GaN and, thus, decreases the density of the two-dimensional electron gas, causing degradation of the channel current. This result leads to the post-etching recovery treatment without a high temperature process and ohmic contact degradation.

ID card recognition is an application scenario in image recognition. As an important identity certificate for citizens, ID card identification and information management has also become a hot issue of concern. Chinese ID card images recognition mainly faces the following problems: Firstly, the shadow grid lines and the noise interference caused by the camera hardware conditions and illumination, together with the useful information in the picture, cause a huge problem of extracting useful information in the ID card. Secondly, the way to extract the foreground part in a complex context is also important for recognition. In response to these problems, this paper proposes an ID card region location method based on face detection and national emblem detection. This method can locate the foreground target area of the ID card in a complex context. In addition, the method of rotation correction of ID card region based on Hough transform is proposed, which can correct the tilt target within the range of tilt angle less than 45 degrees. In the text localization stage, the morphological method of the image is used to locate the text area. The deep convolutional neural network is then used to implement character segmentation and recognition. The recognition process proposed in this paper achieves a better recognition effect, and the recognition of Chinese and digital characters reaches 99.7% recognition rate, and can cope with the rotation tilt and the image containing the background.

Gallium Nitride (GaN) high electron mobility transistors (HEMTs) have become good candidate for microwave power amplifier application. Furthermore, by combining the high power and the low noise performance, GaN HEMTs also show a great potential in high linearity low noise amplifier application. In this report, a four-stage low noise amplifier using commercial 100 nm GaN on Silicon Technology is presented. The designed LNA achieves a noise figure of 2.2-2.8 dB and a small signal gain of 25–29 dB from 23 to 31 GHz. The output 3-dB compression point (P3dB) is in the 17 dBm range. The LNA works with only one DC power supply and has a chip size of $2.3\mathbf{mm}\times 1\mathbf{mm}$ , Compared with the existing Gallium Arsenic technology, this LNA has higher gain, higher linearity while keeping the noise figure at the same level.

In this paper, a Ka-band MMIC power amplifier using 100 nm gate-length GaN process is presented. It was designed to get high output power and high gain. The power amplifier provides 7W of output power with a minimum 32% PAE working at 23–30 GHz. We employed 100nm GaN process on 100um thick Si substrate technology. The out power of this MMIC is comparable to that of PA MMIC in GaN process on SiC substrate.

GaN technology has attracted main attention towards its application to high-power amplifier. Most recently, noise performance of GaN device has also won acceptance. Compared with GaAs LNA, GaN LNA has a unique superiority on power handling. In this work, we report a wideband Silicon-substrate GaN MMIC LNA operating in 18–31 GHz frequency range using a commercial 0.1 um T-Gate HEMT process. The GaN MMIC LNA has an average noise figure of 1.43 dB over the band and a minimum value of 1.27 dB at 23.2 GHz, which can compete with GaAs and InP MMIC LNA. The small-signal gain is between 22 and 25 dB across the band, the input and output return losses of the MMIC are less than −10 dB. The P1dB and OIP3 are at 17 dBm and 28 dBm level. The four-stage MMIC is 2.3×1.0 mm2 in area and consumes 280 mW DC power. Compared with GaAs and InP LNA, the GaN MMIC LNA in this work exhibits a comparative noise figure with higher linearity and power handling ability.

The aim of the present study was to assess the value of susceptibility-weighted imaging (SWI) and diffusion-weighted imaging (DWI) in the grading of gliomas and to evaluate the correlation between these quantitative parameters derived from SWI and DWI. A total of 49 patients with glioma were assessed by DWI and SWI. The evaluation included the ratio of apparent diffuse coefficient values between the solid portion of tumors and contralateral normal white matter (rADC) and the degree of intratumoral susceptibility signal intensity (ITSS) within tumors. Receiver operating characteristic curve (ROC) analyses were performed and the area under the ROC curve was calculated to compare the diagnostic performance, determine optimum thresholds for tumor grading, and calculate the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for identifying high-grade gliomas. The correlation between DWI- and SWI-derived parameters was also evaluated. The rADC and the degrees of ITSS within tumors were significantly higher in high-grade gliomas than those in low-grade gliomas. ROC curve analysis indicated that the rADC was a better index for grading gliomas than the ITSS degree. Statistical analysis demonstrated a threshold value of 1.497 for rADC to provide a sensitivity, specificity, PPV and NPV of 86.2, 85.0, 89.3 and 81.0%, respectively, for determining high-grade gliomas. A degree of ITSS of 1.5 was defined as the threshold to identify high-grade gliomas and sensitivity, specificity, PPV and NPV of 82.8, 75.0, 82.8 and 75.0% were obtained, respectively. Furthermore, a moderate inverse correlation between rADC and the ITSS degree was revealed. Combination of SWI with DWI may provide valuable information for glioma grading.

An optimal design of light-emitting diode (LED) lighting that benefits both the photosynthesis performance for plants and the visional health for human eyes has drawn considerable attention. In the present study, we have developed a multi-color driving algorithm that serves as a liaison between desired spectral power distributions and pulse-width-modulation duty cycles. With the aid of this algorithm, our multi-color plant-growth light sources can optimize correlated-color temperature (CCT) and color rendering index (CRI) such that photosynthetic luminous efficacy of radiation (PLER) is maximized regardless of the number of LEDs and the type of photosynthetic action spectrum (PAS). In order to illustrate the accuracies of the proposed algorithm and the practicalities of our plant-growth light sources, we choose six color LEDs and German PAS for experiments. Finally, our study can help provide a useful guide to improve light qualities in plant factories, in which long-term co-inhabitance of plants and human beings is required.

The origin of the threshold voltage (Vth) shift that occurs in ON-state biased AlGaN/GaN high-electron mobility transistors (HEMTs) is proposed in terms of experimental researches and first-principles calculations. Experimentally, ∼20% negative shift of the Vth is found for the 0.2 μm gate length T-gate AlGaN/GaN HEMT under the ON-state stress. 1/f noise measurements are carried out to investigate the evolution of defect distribution in the GaN channel, which indicates that the Vth shift is accompanied by the diminishing of the defect level of Ev + 0.8 eV. Combined with first-principles studies based on hybrid functionals defect calculations, we find that the defect evolution of O from negatively charged VGa-ON to neutral ON-H gives rise to the decrease in acceptors in the GaN channel and thus the negative shift of Vth. Our work indicates that the strengthening of oxygen-related defects in GaN during material growth and device fabrication would improve the reliability of AlGaN/GaN HEMTs.

The level set methods have provided powerful frameworks for image segmentation. However, to obtain accurate boundaries of the objects, especially when they have weak edges or inhomogeneous intensities, is still a very challenging task. Actually, we have studied the popular existing level set approaches and discovered that they failed to segment the images with weak edges or inhomogeneous intensities in many cases. The weak/blurry edges and inhomogeneous intensities cause uncertainty and fuzziness for segmentation. In this paper, a novel fuzzy level set approach is proposed. At first,S-function based on the maximum fuzzy entropy principle (MEP) is used to map the image from space domain to fuzzy domain. Then, an energy function is formulated according to the differences between the actual and estimated probability densities of the intensities in different regions. A partial differential equation is derived for finding the minimum of the energy function. The proposed approach has been tested on both synthetic images and real images and evaluated by several popular metrics. The experimental results demonstrate that the proposed approach can locate the true object boundaries, even for objects with blurry boundaries, low contrast, and inhomogeneous intensities.