Recently, there is a research trend on ad-hoc microphone arrays. However, most research was conducted on simulated data. Although some data sets were collected with a small number of distributed devices, they were not synchronized which hinders the f
undamental theoretical research to ad-hoc microphone arrays. To address this issue, this paper presents a synchronized speech corpus, named Libri-adhoc40, which collects the replayed Librispeech data from loudspeakers by ad-hoc microphone arrays of 40 strongly synchronized distributed nodes in a real office environment. Besides, to provide the evaluation target for speech frontend processing and other applications, we also recorded the replayed speech in an anechoic chamber. We trained several multi-device speech recognition systems on both the Libri-adhoc40 dataset and a simulated dataset. Experimental results demonstrate the validness of the proposed corpus which can be used as a benchmark to reflect the trend and difference of the models with different ad-hoc microphone arrays. The dataset is online available at https://github.com/ISmallFish/Libri-adhoc40.
Multichannel blind audio source separation aims to recover the latent sources from their multichannel mixtures without supervised information. One state-of-the-art blind audio source separation method, named independent low-rank matrix analysis (ILRM
A), unifies independent vector analysis (IVA) and nonnegative matrix factorization (NMF). However, the spectra matrix produced from NMF may not find a compact spectral basis. It may not guarantee the identifiability of each source as well. To address this problem, here we propose to enhance the identifiability of the source model by a minimum-volume prior distribution. We further regularize a multichannel NMF (MNMF) and ILRMA respectively with the minimum-volume regularizer. The proposed methods maximize the posterior distribution of the separated sources, which ensures the stability of the convergence. Experimental results demonstrate the effectiveness of the proposed methods compared with auxiliary independent vector analysis, MNMF, ILRMA and its extensions.
In recent years, Siamese network based trackers have significantly advanced the state-of-the-art in real-time tracking. However, state-of-the-art Siamese trackers suffer from high memory cost which restricts their applicability in mobile applications
having strict constraints on memory budget. To address this issue, we propose a novel distilled Siamese tracking framework to learn small, fast yet accurate trackers (students), which capture critical knowledge from large Siamese trackers (teachers) by a teacher-students knowledge distillation model. This model is intuitively inspired by a one-teacher vs multi-students learning mechanism, which is the most usual teaching method in the school. In particular, it contains a single teacher-student distillation model and a student-student knowledge sharing mechanism. The first one is designed by a tracking-specific distillation strategy to transfer knowledge from the teacher to students. The later is utilized for mutual learning between students to enable an in-depth knowledge understanding. To the best of our knowledge, we are the first to investigate knowledge distillation for Siamese trackers and propose a distilled Siamese tracking framework. We demonstrate the generality and effectiveness of our framework by conducting a theoretical analysis and extensive empirical evaluations on several popular Siamese trackers. The results on five tracking benchmarks clearly show that the proposed distilled trackers achieve compression rates up to 18$times$ and frame-rates of $265$ FPS with speedups of 3$times$, while obtaining similar or even slightly improved tracking accuracy.
In this paper, we define the induced modules of Lie algebra ad$(B)$ associated with a 3-Lie algebra $B$-module, and study the relation between 3-Lie algebra $A_{omega}^{delta}$-modules and induced modules of inner derivation algebra ad$(A_{omega}^{de
lta})$. We construct two infinite dimensional intermediate series modules of 3-Lie algebra $A_{omega}^{delta}$, and two infinite dimensional modules $(V, psi_{lambdamu})$ and $(V, phi_{mu})$ of the Lie algebra ad$(A_{omega}^{delta})$, and prove that only $(V, psi_{lambda0})$ and $(V, psi_{lambda1})$ are induced modules.
Atomic Fermi gases provide an ideal platform for studying the pairing and superfluid physics, using a Feshbach resonance between closed channel molecular states and open channel scattering states. Of particular interest is the strongly interacting re
gime. We show that the closed-channel fraction $Z$ provides an effective probe for the important many-body interacting effects, especially through its density dependence, which is absent from two-body theoretical predictions. Here we measure $Z$ as a function of interaction strength and the Fermi temperature $T_F$ in a trapped $^6$Li superfluid throughout the entire BCS--BEC crossover. Away from the deep BEC regime, the fraction $Z$ is sensitive to $T_F$. In particular, our data show $Z propto T_F^{alpha}$ with $alpha=1/2$ at unitarity, in quantitative agreement with calculations of a two-channel pairing fluctuation theory, and $alpha$ increases rapidly into the BCS regime, reflecting many-body interaction effects as predicted.
Kinetic approaches, i.e., methods based on the lattice Boltzmann equations, have long been recognized as an appealing alternative for solving incompressible Navier-Stokes equations in computational fluid dynamics. However, such approaches have not be
en widely adopted in graphics mainly due to the underlying inaccuracy, instability and inflexibility. In this paper, we try to tackle these problems in order to make kinetic approaches practical for graphical applications. To achieve more accurate and stable simulations, we propose to employ the non-orthogonal central-moment-relaxation model, where we develop a novel adaptive relaxation method to retain both stability and accuracy in turbulent flows. To achieve flexibility, we propose a novel continuous-scale formulation that enables samples at arbitrary resolutions to easily communicate with each other in a more continuous sense and with loose geometrical constraints, which allows efficient and adaptive sample construction to better match the physical scale. Such a capability directly leads to an automatic sample construction which generates static and dynamic scales at initialization and during simulation, respectively. This effectively makes our method suitable for simulating turbulent flows with arbitrary geometrical boundaries. Our simulation results with applications to smoke animations show the benefits of our method, with comparisons for justification and verification.
Recording atomic-resolution transmission electron microscopy (TEM) images is becoming increasingly routine. A new bottleneck is then analyzing this information, which often involves time-consuming manual structural identification. We have developed a
deep learning-based algorithm for recognition of the local structure in TEM images, which is stable to microscope parameters and noise. The neural network is trained entirely from simulation but is capable of making reliable predictions on experimental images. We apply the method to single sheets of defected graphene, and to metallic nanoparticles on an oxide support.
The interaction force between likely charged particles/surfaces is usually repulsive due to the Coulomb interaction. However, the counterintuitive like-charge attraction in electrolytes has been frequently observed in experiments, which has been theo
retically debated for a long time. It is widely known that the mean field Poisson-Boltzmann theory cannot explain or predict this anomalous feature since it ignores many-body properties. In this paper, we develop efficient algorithm and perform the force calculation between two interfaces using a set of self-consistent equations which properly takes into account the electrostatic correlation and the dielectric-boundary effects. By solving the equations and calculating the pressure with the Debye-charging process, we show that the self-consistent equations could be used to study the attraction between like-charge surfaces from weak-coupling to mediate-coupling regime, and that the attraction is due to the electrostatics-driven entropic force which is significantly enhanced by the dielectric depletion of mobile ions. A systematic investigation shows that the interaction forces can be tuned by material permittivity, ionic size and valence, and salt concentration, and that the like-charge attraction exists only for specific regime of these parameters.
We propose a modified Poisson-Nernst-Planck (PNP) model to investigate charge transport in electrolytes of inhomogeneous dielectric environment. The model includes the ionic polarization due to the dielectric inhomogeneity and the ion-ion correlation
. This is achieved by the self energy of test ions through solving a generalized Debye-Huckel (DH) equation. We develop numerical methods for the system composed of the PNP and DH equations. Particularly, towards the numerical challenge of solving the high-dimensional DH equation, we developed an analytical WKB approximation and a numerical approach based on the selective inversion of sparse matrices. The model and numerical methods are validated by simulating the charge diffusion in electrolytes between two electrodes, for which effects of dielectrics and correlation are investigated by comparing the results with the prediction by the classical PNP theory. We find that, at the length scale of the interface separation comparable to the Bjerrum length, the results of the modified equations are significantly different from the classical PNP predictions mostly due to the dielectric effect. It is also shown that when the ion self energy is in weak or mediate strength, the WKB approximation presents a high accuracy, compared to precise finite-difference results.
Relay-assisted cooperative wireless communication has been shown to have significant performance gains over the legacy direct transmission scheme. Compared with single relay based cooperation schemes, utilizing multiple relays further improves the re
liability and rate of transmissions. Distributed space-time coding (DSTC), as one of the schemes to utilize multiple relays, requires tight coordination between relays and does not perform well in a distributed environment with mobility. In this paper, a cooperative medium access control (MAC) layer protocol, called emph{STiCMAC}, is designed to allow multiple relays to transmit at the same time in an IEEE 802.11 network. The transmission is based on a novel DSTC scheme called emph{randomized distributed space-time coding} (emph{R-DSTC}), which requires minimum coordination. Unlike conventional cooperation schemes that pick nodes with good links, emph{STiCMAC} picks a emph{transmission mode} that could most improve the end-to-end data rate. Any station that correctly receives from the source can act as a relay and participate in forwarding. The MAC protocol is implemented in a fully decentralized manner and is able to opportunistically recruit relays on the fly, thus making it emph{robust} to channel variations and user mobility. Simulation results show that the network capacity and delay performance are greatly improved, especially in a mobile environment.
