arXiv subjects

Ye Li

Publications and source records attributed to Ye Li.

At least 199 records · Page 11Linked to original sources

arXiv2020

Generating Anthropomorphic Phantoms Using Fully Unsupervised Deformable Image Registration with Convolutional Neural Networks

Objectives: Computerized phantoms play an essential role in various applications of medical imaging research. Although the existing computerized phantoms can model anatomical variations through organ and phantom scaling, this does not provide a way to fully reproduce anatomical variations seen in humans. However, having a population of phantoms that models the variations in patient anatomy and, in nuclear medicine, uptake realization is essential for comprehensive validation and training. In this work, we present a novel image registration method for creating highly anatomically detailed anthropomorphic phantoms from a single digital phantom. Methods: We propose a deep-learning-based registration algorithm to predict deformation parameters for warping an XCAT phantom to a patient CT scan. This proposed algorithm optimizes a novel SSIM-based objective function for a given image pair independently of the training data and thus is truly and fully unsupervised. We evaluate the proposed method on a publicly available low-dose CT dataset from TCIA. Results: The performance of the proposed model was compared with that of several state-of-the-art methods, and outperformed them by more than 8%, measured by the SSIM and less than 30%, by the MSE. Conclusion: A deep-learning-based unsupervised registration method was developed to create anthropomorphic phantoms while providing "gold-standard" anatomies that can be used as the basis for modeling organ properties. Significance: Experimental results demonstrate the effectiveness of the proposed method. The resulting anthropomorphic phantom is highly realistic. Combined with realistic simulations of the image formation process, the generated phantoms could serve in many applications of medical imaging research.

eess.IV↗

arXiv2020

Triplet Online Instance Matching Loss for Person Re-identification

Mining the shared features of same identity in different scene, and the unique features of different identity in same scene, are most significant challenges in the field of person re-identification (ReID). Online Instance Matching (OIM) loss function and Triplet loss function are main methods for person ReID. Unfortunately, both of them have drawbacks. OIM loss treats all samples equally and puts no emphasis on hard samples. Triplet loss processes batch construction in a complicated and fussy way and converges slowly. For these problems, we propose a Triplet Online Instance Matching (TOIM) loss function, which lays emphasis on the hard samples and improves the accuracy of person ReID effectively. It combines the advantages of OIM loss and Triplet loss and simplifies the process of batch construction, which leads to a more rapid convergence. It can be trained on-line when handle the joint detection and identification task. To validate our loss function, we collect and annotate a large-scale benchmark dataset (UESTC-PR) based on images taken from surveillance cameras, which contains 499 identities and 60,437 images. We evaluated our proposed loss function on Duke, Marker-1501 and UESTC-PR using ResNet-50, and the result shows that our proposed loss function outperforms the baseline methods by a maximum of 21.7%, including Softmax loss, OIM loss and Triplet loss.

cs.CV↗

arXiv2020

Acoustic Scene Classification Using Bilinear Pooling on Time-liked and Frequency-liked Convolution Neural Network

The current methodology in tackling Acoustic Scene Classification (ASC) task can be described in two steps, preprocessing of the audio waveform into log-mel spectrogram and then using it as the input representation for Convolutional Neural Network (CNN). This paradigm shift occurs after DCASE 2016 where this framework model achieves the state-of-the-art result in ASC tasks on the (ESC-50) dataset and achieved an accuracy of 64.5%, which constitute to 20.5% improvement over the baseline model, and DCASE 2016 dataset with an accuracy of 90.0% (development) and 86.2% (evaluation), which constitute a 6.4% and 9% improvements with respect to the baseline system. In this paper, we explored the use of harmonic and percussive source separation (HPSS) to split the audio into harmonic audio and percussive audio, which has received popularity in the field of music information retrieval (MIR). Although works have been done in using HPSS as input representation for CNN model in ASC task, this paper further investigate the possibility on leveraging the separated harmonic component and percussive component by curating 2 CNNs which tries to understand harmonic audio and percussive audio in their natural form, one specialized in extracting deep features in time biased domain and another specialized in extracting deep features in frequency biased domain, respectively. The deep features extracted from these 2 CNNs will then be combined using bilinear pooling. Hence, presenting a two-stream time and frequency CNN architecture approach in classifying acoustic scene. The model is being evaluated on DCASE 2019 sub task 1a dataset and scored an average of 65% on development dataset, Kaggle Leadership Private and Public board.

eess.AS↗

arXiv2020

Non-Negative Matrix Factorization-Convolutional Neural Network (NMF-CNN) For Sound Event Detection

The main scientific question of this year DCASE challenge, Task 4 - Sound Event Detection in Domestic Environments, is to investigate the types of data (strongly labeled synthetic data, weakly labeled data, unlabeled in domain data) required to achieve the best performing system. In this paper, we proposed a deep learning model that integrates Non-Negative Matrix Factorization (NMF) with Convolutional Neural Network (CNN). The key idea of such integration is to use NMF to provide an approximate strong label to the weakly labeled data. Such integration was able to achieve a higher event-based F1-score as compared to the baseline system (Evaluation Dataset: 30.39% vs. 23.7%, Validation Dataset: 31% vs. 25.8%). By comparing the validation results with other participants, the proposed system was ranked 8th among 19 teams (inclusive of the baseline system) in this year Task 4 challenge.

cs.SD↗

arXiv2019

A unified binary neutron star merger magnetar model for the Chandra X-ray transients CDF-S XT1 and XT2

Two bright X-ray transients were reported from the Chandra Deep Field South archival data, namely CDF-S XT1 and XT2. Whereas the nature of the former is not identified, the latter was suggested as an excellent candidate for a rapidly spinning magnetar born from a binary neutron star (BNS) merger. Here we propose a unified model to interpret both transients within the framework of the BNS merger magnetar model. According to our picture, CDF-S XT2 is observed from the "free zone" where the magnetar spindown powered X-ray emission escapes freely, whereas CDF-S XT1 originates from the "trapped zone" where the X-ray emission is initially blocked by the dynamical ejecta and becomes transparent after the ejecta is pushed to a distance where Thomson optical depth drops below unity. We fit the magnetar model to the light curves of both transients and derived consistent parameters for the two events, with magnetic field, initial spin period and X-ray emission efficiency being ($B_p=10^{16}\,G$, $P=1.2\,\rm ms$, $η= 0.001$) and ($B_p=10^{15.8}\,G$, $P=4.4\, \rm ms$, $η= 0.001$) for XT1 and XT2, respectively. The "isotropic equivalent" ejecta mass of XT1 is $M_{\rm ej} \sim 10^{-3}$ $M_{\odot}$, while it is not constrained for XT2. Our results suggest that more extreme magnetar parameters are required to have XT1 detected from the trapped zone. The model parameters for both events are generally consistent with those derived from SGRB X-ray plateau observations. The host galaxy properties of both transients are also consistent with those of SGRBs. The event rate densities of both XT1 and XT2 are consistent with that of BNS mergers.

astro-ph.HE↗

arXiv2019

Non-detection of fast radio bursts from six gamma-ray burst remnants with possible magnetar engines

The analogy of the host galaxy of the repeating fast radio burst (FRB) source FRB 121102 and those of long gamma-ray bursts (GRBs) and super-luminous supernovae (SLSNe) has led to the suggestion that young magnetars born in GRBs and SLSNe could be the central engine of repeating FRBs. We test such a hypothesis by performing dedicated observations of the remnants of six GRBs with evidence of having a magnetar central engine using the Arecibo telescope and the Robert C. Byrd Green Bank Telescope (GBT). A total of $\sim 20$ hrs of observations of these sources did not detect any FRB from these remnants. Under the assumptions that all these GRBs left behind a long-lived magnetar and that the bursting rate of FRB 121102 is typical for a magnetar FRB engine, we estimate a non-detection probability of $8.9\times10^{-6}$. Even though these non-detections cannot exclude the young magnetar model of FRBs, we place constraints on the burst rate and luminosity function of FRBs from these GRB targets.

astro-ph.HE↗

arXiv2019

The FRB 121102 host is atypical among nearby FRBs

We search for host galaxy candidates of nearby fast radio bursts (FRBs), FRB 180729.J1316+55, FRB 171020, FRB 171213, FRB 180810.J1159+83, and FRB 180814.J0422+73 (the second repeating FRB). We compare the absolute magnitudes and the expected host dispersion measure $\rm DM_{host}$ of these candidates with that of the first repeating FRB, FRB 121102, as well as those of long gamma ray bursts (LGRBs) and superluminous supernovae (SLSNe), the proposed progenitor systems of FRB 121102. We find that while the FRB 121102 host is consistent with those of LGRBs and SLSNe, the nearby FRB host candidates, at least for FRB 180729.J1316+55, FRB 171020, and FRB180814.J0422+73, either have a smaller $\rm DM_{host}$ or are fainter than FRB121102 host, as well as the hosts of LGRBs and SLSNe. In order to avoid the uncertainty in estimating $\rm DM_{host}$ due to the line-of-sight effect, we propose a galaxy-group-based method to estimate the electron density in the inter-galactic regions, and hence, $\rm DM_{IGM}$. The result strengthens our conclusion. We conclude that the host galaxy of FRB 121102 is atypical, and LGRBs and SLSNe are likely not the progenitor systems of at least most nearby FRB sources. {The recently reported two FRB hosts differ from the host of FRB 121102 and also the host candidates suggested in this paper. This is consistent with the conclusion of our paper and suggests that the FRB hosts are very diverse. }

astro-ph.HE↗

arXiv2019

Wearable Travel Aid for Environment Perception and Navigation of Visually Impaired People

This paper presents a wearable assistive device with the shape of a pair of eyeglasses that allows visually impaired people to navigate safely and quickly in unfamiliar environment, as well as perceive the complicated environment to automatically make decisions on the direction to move. The device uses a consumer Red, Green, Blue and Depth (RGB-D) camera and an Inertial Measurement Unit (IMU) to detect obstacles. As the device leverages the ground height continuity among adjacent image frames, it is able to segment the ground from obstacles accurately and rapidly. Based on the detected ground, the optimal walkable direction is computed and the user is then informed via converted beep sound. Moreover, by utilizing deep learning techniques, the device can semantically categorize the detected obstacles to improve the users' perception of surroundings. It combines a Convolutional Neural Network (CNN) deployed on a smartphone with a depth-image-based object detection to decide what the object type is and where the object is located, and then notifies the user of such information via speech. We evaluated the device's performance with different experiments in which 20 visually impaired people were asked to wear the device and move in an office, and found that they were able to avoid obstacle collisions and find the way in complicated scenarios.

cs.CV↗

arXiv2019

Multimessenger Tests of Einstein's Weak Equivalence Principle and Lorentz Invariance with a High-energy Neutrino from a Flaring Blazar

The detection of the high-energy ($\sim290$ TeV) neutrino coincident with the flaring blazar TXS 0506+056, the first and only $3σ$ neutrino-source association to date, provides new, multimessenger tests of the weak equivalence principle (WEP) and Lorentz invariance. Assuming that the flight time difference between the TeV neutrino and gamma-ray photons from the blazar flare is mainly caused by the gravitational potential of the Laniakea supercluster of galaxies, we show that the deviation from the WEP for neutrinos and photons is conservatively constrained to have an accuracy of $10^{-6}-10^{-7}$, which is 3--4 orders of magnitude better than previous results placed by MeV neutrinos from supernova 1987A. In addition, we demonstrate that the association of the TeV neutrino with the blazar flare sets limits on the energy scales of quantum gravity for both linear and quadratic violations of Lorentz invariance (LIV) to $E_{\rm QG, 1}>3.2\times10^{15}-3.7\times10^{16}$ GeV and $E_{\rm QG, 2}>4.0\times10^{10}-1.4\times10^{11}$ GeV. These improve previous limits on both linear and quadratic LIV energy scales in neutrino propagation by 5--7 orders of magnitude.

astro-ph.HE↗

arXiv2018

Precise QCD Description of the Higgs Boson Transverse Momentum Spectrum

The transverse momentum ($p_T$) distribution of Higgs bosons at hadron colliders enables a detailed probe of its production dynamics and is a key ingredient to precision studies of Higgs boson properties, but receives very large QCD corrections. We obtain a precision prediction for the $p_T$ spectrum by matching second-order (NNLO) QCD corrections at large $p_T$ with resummation of third-order logarithmic (N$^3$LL) corrections at small $p_T$. We achieve significantly improved results for $p_T< 35\,{\rm GeV}$ with perturbative uncertainties $\lesssim \pm 6\%$, and thus a convergent perturbative series for all values of $p_T$.

hep-ph↗

arXiv2018

Hadronic Final States in DIS at NNLO QCD with Parton Showers

We present a parton-shower matched NNLO QCD calculation for hadronic final state production in Deep Inelastic Scattering. The computation is based on the UNLOPS method and is implemented in the publicly available event generation framework SHERPA. Results are compared to measurements performed by the H1 collaboration.

hep-ph↗

arXiv2018

Bootstrapping rapidity anomalous dimension for transverse-momentum resummation

Soft function relevant for transverse-momentum resummation for Drell-Yan or Higgs production at hadron colliders are computed through to three loops in the expansion of strong coupling, with the help of bootstrap technique and supersymmetric decomposition. The corresponding rapidity anomalous dimension is extracted. An intriguing relation between anomalous dimensions for transverse-momentum resummation and threshold resummation is found.

hep-ph↗

arXiv2018

SDSS J211852.96-073227.5: a new γ-ray flaring narrow-line Seyfert 1 galaxy

We report on the identification of a new γ-ray-emitting narrow-line Seyfert 1 (NLS1) galaxy, SDSS J211852.96-073227.5 (hereinafter J2118-0732). The galaxy, at a redshift of 0.26, is associated with a radio source of flat/inverted spectrum at high radio frequencies. The analysis of its optical spectrum obtained in the Sloan Digital Sky Survey (SDSS) revealed a small linewidth of the broad component of the Hβ line (full width at half-maximum = 1585 km s^{-1}), making it a radio-loud NLS1 galaxy -- an intriguing class of active galactic nuclei with exceptional multiwavelength properties. A new γ-ray source centred at J2118-0732 was sporadically detected during 2009--2013 in form of flares by the Fermi-LAT. Our XMM-Newton observations revealed a flat X-ray spectrum described by a simple power law, and a flux variation by a factor of ~2.5 in five months. The source also shows intraday variability in the infrared band. Its broad-band spectral energy distribution can be modelled by emission from a simple one-zone leptonic jet model, and the flux drop from infrared to X-rays in five months can be explained by changes of the jet parameters, though the exact values may be subject to relatively large uncertainties. With the NLS1-blazar composite nucleus, the clear detection of the host galaxy, and the synchronous variations in the multiwavelength fluxes, J2118-0732 provides a new perspective on the formation and evolution of relativistic jets under the regime of relatively small black hole masses and high accretion rates.

astro-ph.HE↗

arXiv2018

Are there multiple populations of Fast Radio Bursts?

The repeating FRB 121102 (the "repeater") shows {repetitive bursting activities} and was localized in a host galaxy at $z=0.193$. On the other hand, despite dozens of hours of telescope time spent on follow-up observations, no other FRBs have been observed to repeat. Yet, it has been speculated that the repeater is the prototype of FRBs, and that other FRBs should show similar repeating patterns. Using the published data, we compare the repeater with other FRBs in the observed time interval ($Δt$) - flux ratio ($S_i / S_{i+1}$) plane. We find that whereas other FRBs occupy the upper (large $S_i / S_{i+1}$) and right (large $Δt$) regions of the plane due to the non-detections of other bursts, some of the repeater bursts fall into the lower-left region of the plot (short interval and small flux ratio) excluded by the non-detection data of other FRBs. The trend also exists even if one only selects those bursts detectable by the Parkes radio telescope. If other FRBs were similar to the repeater, our simulations suggest that the probability that none of them have been detected to repeat with the current searches would be $\sim (10^{-4}-10^{-3})$. We suggest that the repeater is not representative of the entire FRB population and that there is strong evidence of more than one population of FRBs.

astro-ph.HE↗

arXiv2018

Fermi Large Area Telescope detection of gamma-ray emission from the direction of supernova iPTF14hls

The remnant of supernova explosion is widely believed to be the acceleration site of high-energy cosmic ray particles. The acceleration timescale is, however, typically very long. Here we report the detection of a variable $γ$-ray source with the Fermi Large Area Telescope, which is positionally and temporally consistent with a peculiar supernova, iPTF14hls. A quasi-stellar object SDSS J092054.04+504251.5, which is probably a blazar according to the infrared data, is found in the error circle of the $γ$-ray source. More data about the $γ$-ray source and SDSS J092054.04+504251.5 are needed to confirm their association. On the other hand, if the association between the $γ$-ray source and the supernova is confirmed, this would be the first time to detect high-energy $γ$-ray emission from a supernova, suggesting very fast particle acceleration by supernova explosions.

astro-ph.HE↗

arXiv2018

Effects of blockage, arrangement and channel dynamics on performance of turbines in a tidal array

The performance and economics of turbines in a tidal array are largely dependent on the power per turbine, and so approaches that can increase this power are crucial for the development of tidal energy. In this paper, we combine a two-scale partial array model and a one-dimensional channel model to investigate the effects of blockage, turbine arrangement, and channel dynamics on tidal turbines. The power per turbine is obtained as the product of two parameters: a power coefficient measuring the power acquired from the instantaneous flow and an environment coefficient showing the response of the channel to added drag. The results suggest that taking account of channel dynamics will decrease the predicted power and the optimal induction factor. The model also shows that when the number of turbines in a row is increased, the power per turbine may monotonically increase or decrease or attain a maximum value at a certain global blockage. These different results depend on two characteristic parameters of the channel: $α$ and $λ_D$. Furthermore, we find that besides turbine density (blockage), the arrangement of the turbines should also be considered if we want to obtain an efficient array. Appropriate arrangements can enhance the performance of turbines in tidal channels, although the beneficial effects will be partly offset by reduced velocities. The turbine arrangement also has an effect on the optimal global blockage at which the power attains its maximum value. As the rate of increase of the power per turbine from an array spanning the whole channel width to the optimal partial array diminishes with increasing blockage, the optimal global blockage will also decrease.

physics.flu-dyn↗

arXiv2017

Performance of two-dimensional tidal turbine arrays in free surface flow

Encouraged by recent studies on the performance of tidal turbine arrays, we extend the classical momentum actuator disc theory to include the free surface effects and allow the vertical arrangement of turbines. Most existing literatures concern one dimensional arrays with single turbine in the vertical direction, while the arrays in this work are two dimensional (with turbines in both the vertical and lateral directions) and also partially block the channel which width is far larger than height. The vertical mixing of array scale flow is assumed to take place much faster than lateral one. This assumption has been verified by numerical simulations. Fixing the total turbine area and utilized width, the comparison between two-dimensional and traditional one-dimensional arrays is investigated. The results suggest that the two dimensional arrangements of smaller turbines are preferred to one dimensional arrays from both the power coefficient and efficiency perspectives. When channel dynamics are considered, the power increase would be partly offset according to the parameters of the channel but the optimal arrangement is unchangeable. Furthermore, we consider how to arrange finite number of turbines in a channel. It is shown that an optimal distribution of turbines in two directions is found. Finally, the scenario of arranging turbines in infinite flow, which is the limiting condition of small blockages, is analysed. A new maximum power coefficient 0.869 occurs when $Fr=0.2$, greatly increasing the peak power compared with existing results.

physics.flu-dyn↗

arXiv2017

Searching for magnetar powered merger-novae from short GRBs

The merger of a double neutron star (NS-NS) binary may result in a rapidly rotating massive NS with an extremely strong magnetic field (i.e., a millisecond magnetar). In this case, the magnetic spin-down of the NS remnant provides an additional source of sustained energy injection, which would continuously power the merger ejecta. The thermal emission from the merger ejecta would give rise to a bright optical "magnetar-powered merger-nova". In this work, we carry out a complete search for magnetar-powered merger-nova from Swift short gamma-ray burst (SGRB) sample. We focus on short GRBs with extended emission or internal plateau, which may signify the presence of magnetars as the central engine. We eventually find three candidates of magnetar-powered merger-nova from the late observations of GRB 050724, GRB 070714B and GRB 061006. With standard parameter values, the magnetar remnant scenario could well interpret the multi-band data of all three bursts, including the extended emission and their late chromatic features in the optical and X-ray data. The peak luminosities of these merger-novae reach several times $10^{42}~{\rm erg~s^{-1}}$, more than one order of magnitude brighter than the traditional "kilo-novae" with peak luminosity of $\sim 10^{41}~{\rm erg~s^{-1}}$. Intense, multi-color late time observations of short GRBs are encouraged to identify more merger-novae in the future.

astro-ph.HE↗