arXiv subjects

Dong Wang

Publications and source records attributed to Dong Wang.

At least 739 records · Page 41Linked to original sources

arXiv2017

Speaker Recognition with Cough, Laugh and "Wei"

This paper proposes a speaker recognition (SRE) task with trivial speech events, such as cough and laugh. These trivial events are ubiquitous in conversations and less subjected to intentional change, therefore offering valuable particularities to discover the genuine speaker from disguised speech. However, trivial events are often short and idiocratic in spectral patterns, making SRE extremely difficult. Fortunately, we found a very powerful deep feature learning structure that can extract highly speaker-sensitive features. By employing this tool, we studied the SRE performance on three types of trivial events: cough, laugh and "Wei" (a short Chinese "Hello"). The results show that there is rich speaker information within these trivial events, even for cough that is intuitively less speaker distinguishable. With the deep feature approach, the EER can reach 10%-14% with the three trivial events, despite their extremely short durations (0.2-1.0 seconds).

cs.SD↗

arXiv2017

Cross-lingual Speaker Verification with Deep Feature Learning

Existing speaker verification (SV) systems often suffer from performance degradation if there is any language mismatch between model training, speaker enrollment, and test. A major cause of this degradation is that most existing SV methods rely on a probabilistic model to infer the speaker factor, so any significant change on the distribution of the speech signal will impact the inference. Recently, we proposed a deep learning model that can learn how to extract the speaker factor by a deep neural network (DNN). By this feature learning, an SV system can be constructed with a very simple back-end model. In this paper, we investigate the robustness of the feature-based SV system in situations with language mismatch. Our experiments were conducted on a complex cross-lingual scenario, where the model training was in English, and the enrollment and test were in Chinese or Uyghur. The experiments demonstrated that the feature-based system outperformed the i-vector system with a large margin, particularly with language mismatch between enrollment and test.

cs.SD↗

arXiv2017

Factor Models for Matrix-Valued High-Dimensional Time Series

In finance, economics and many other fields, observations in a matrix form are often observed over time. For example, many economic indicators are obtained in different countries over time. Various financial characteristics of many companies are reported over time. Although it is natural to turn a matrix observation into a long vector then use standard vector time series models or factor analysis, it is often the case that the columns and rows of a matrix represent different sets of information that are closely interrelated in a very structural way. We propose a novel factor model that maintains and utilizes the matrix structure to achieve greater dimensional reduction as well as finding clearer and more interpretable factor structures. Estimation procedure and its theoretical properties are investigated and demonstrated with simulated and real examples.

stat.ME↗

arXiv2017

Critical Height of the Torus Instability in Two-Ribbon Solar Flares

We studied the background field for 60 two-ribbon flares of M-and-above classes during 2011--2015. These flares are categorized into two groups, i.e., \emph{eruptive} and \emph{confined} flares, based on whether a flare is associated with a coronal mass ejection or not. The background field of source active regions is approximated by a potential field extrapolated from the $B_z$ component of vector magnetograms provided by the Helioseismic and Magnetic Imager. We calculated the decay index $n$ of the background field above the flaring polarity inversion line, and defined a critical height $h_\mathrm{crit}$ corresponding to the theoretical threshold ($n_\mathrm{crit}=1.5$) of the torus instability. We found that $h_\mathrm{crit}$ is approximately half of the distance between the centroids of opposite polarities in active regions, and that the distribution of $h_\mathrm{crit}$ is bimodal: it is significantly higher for confined flares than for eruptive ones. The decay index increases monotonously with increasing height for 86\% (84\%) of the eruptive (confined) flares but displays a saddle-like profile for the rest 14\% (16\%), which are found exclusively in active regions of multipolar field configuration. Moreover, $n$ at the saddle bottom is significantly smaller in confined flares than that in eruptive ones. These results highlight the critical role of background field in regulating the eruptive behavior of two-ribbon flares.

astro-ph.SR↗

arXiv2017

A Study on Replay Attack and Anti-Spoofing for Automatic Speaker Verification

For practical automatic speaker verification (ASV) systems, replay attack poses a true risk. By replaying a pre-recorded speech signal of the genuine speaker, ASV systems tend to be easily fooled. An effective replay detection method is therefore highly desirable. In this study, we investigate a major difficulty in replay detection: the over-fitting problem caused by variability factors in speech signal. An F-ratio probing tool is proposed and three variability factors are investigated using this tool: speaker identity, speech content and playback & recording device. The analysis shows that device is the most influential factor that contributes the highest over-fitting risk. A frequency warping approach is studied to alleviate the over-fitting problem, as verified on the ASV-spoof 2017 database.

cs.SD↗

arXiv2017

Weakly Supervised PLDA Training

PLDA is a popular normalization approach for the i-vector model, and it has delivered state-of-the-art performance in speaker verification. However, PLDA training requires a large amount of labelled development data, which is highly expensive in most cases. We present a cheap PLDA training approach, which assumes that speakers in the same session can be easily separated, and speakers in different sessions are simply different. This results in `weak labels' which are not fully accurate but cheap, leading to a weak PLDA training. Our experimental results on real-life large-scale telephony customer service achieves demonstrated that the weak training can offer good performance when human-labelled data are limited. More interestingly, the weak training can be employed as a discriminative adaptation approach, which is more efficient than the prevailing unsupervised method when human-labelled data are insufficient.

cs.LG↗

arXiv2017

Collaborative Learning for Language and Speaker Recognition

This paper presents a unified model to perform language and speaker recognition simultaneously and altogether. The model is based on a multi-task recurrent neural network where the output of one task is fed as the input of the other, leading to a collaborative learning framework that can improve both language and speaker recognition by borrowing information from each other. Our experiments demonstrated that the multi-task model outperforms the task-specific models on both tasks.

cs.SD↗

arXiv2017

Phone-aware Neural Language Identification

Pure acoustic neural models, particularly the LSTM-RNN model, have shown great potential in language identification (LID). However, the phonetic information has been largely overlooked by most of existing neural LID models, although this information has been used in the conventional phonetic LID systems with a great success. We present a phone-aware neural LID architecture, which is a deep LSTM-RNN LID system but accepts output from an RNN-based ASR system. By utilizing the phonetic knowledge, the LID performance can be significantly improved. Interestingly, even if the test language is not involved in the ASR training, the phonetic knowledge still presents a large contribution. Our experiments conducted on four languages within the Babel corpus demonstrated that the phone-aware approach is highly effective.

cs.CL↗

arXiv2017

Muttalib--Borodin ensembles in random matrix theory --- realisations and correlation functions

Muttalib--Borodin ensembles are characterised by the pair interaction term in the eigenvalue probability density function being of the form $\prod_{1 \le j < k \le N}(λ_k - λ_j) (λ_k^θ- λ_j^θ)$. We study the Laguerre and Jacobi versions of this model --- so named by the form of the one-body interaction terms --- and show that for $θ\in \mathbb Z^+$ they can be realised as the eigenvalue PDF of certain random matrices with Gaussian entries. For general $θ> 0$, realisations in terms of the eigenvalue PDF of ensembles involving triangular matrices are given. In the Laguerre case this is a recent result due to Cheliotis, although our derivation is different. We make use of a generalisation of a double contour integral formula for the correlation functions contained in a paper by Adler, van Moerbeke and Wang to analyse the global density (which we also analyse by studying characteristic polynomials), and the hard edge scaled correlation functions. For the global density functional equations for the corresponding resolvents are obtained; solving this gives the moments in terms of Fuss--Catalan numbers (Laguerre case --- a known result) and particular binomial coefficients (Jacobi case). For $θ\in \mathbb Z^+$ the Laguerre and Jacobi cases are closely related to the squared singular values for products of $θ$ standard Gaussian random matrices, and truncations of unitary matrices, respectively. At the hard edge the double contour integral formulas provide a double contour integral form of the scaled correlation kernel obtained by Borodin in terms of Wright's Bessel function.

math-ph↗

arXiv2017

Deep Speaker Feature Learning for Text-independent Speaker Verification

Recently deep neural networks (DNNs) have been used to learn speaker features. However, the quality of the learned features is not sufficiently good, so a complex back-end model, either neural or probabilistic, has to be used to address the residual uncertainty when applied to speaker verification, just as with raw features. This paper presents a convolutional time-delay deep neural network structure (CT-DNN) for speaker feature learning. Our experimental results on the Fisher database demonstrated that this CT-DNN can produce high-quality speaker features: even with a single feature (0.3 seconds including the context), the EER can be as low as 7.68%. This effectively confirmed that the speaker trait is largely a deterministic short-time property rather than a long-time distributional pattern, and therefore can be extracted from just dozens of frames.

cs.SD↗

arXiv2017

Flexible and Creative Chinese Poetry Generation Using Neural Memory

It has been shown that Chinese poems can be successfully generated by sequence-to-sequence neural models, particularly with the attention mechanism. A potential problem of this approach, however, is that neural models can only learn abstract rules, while poem generation is a highly creative process that involves not only rules but also innovations for which pure statistical models are not appropriate in principle. This work proposes a memory-augmented neural model for Chinese poem generation, where the neural model and the augmented memory work together to balance the requirements of linguistic accordance and aesthetic innovation, leading to innovative generations that are still rule-compliant. In addition, it is found that the memory mechanism provides interesting flexibility that can be used to generate poems with different styles.

cs.AI↗

arXiv2017

Exploration quantum steering, nonlocality and entanglement of two-qubit X-state in structured reservoirs

In this work, there are two parties, Alice on Earth and Bob on the satellite, which initially share an entangled state, and some open problems, which emerge during quantum steering that Alice remotely steers Bob, are investigated. Our analytical results indicate that all entangled pure states and maximally entangled evolution states (EESs) are steerable, and not every entangled evolution state is steerable and some steerable states are only locally correlated. Besides, quantum steering from Alice to Bob experiences a "sudden death" with increasing decoherence strength. However, shortly after that, quantum steering experiences a recovery with the increase of decoherence strength in bit flip (BF) and phase flip (PF) channels. Interestingly, while they initially share an entangled pure state, all EESs are steerable and obey Bell nonlocality in PF and phase damping channels. In BF channels, all steerable states can violate Bell-CHSH inequality, but some EESs are unable to be employed to realize steering. However, when they initially share an entangled mixed state, the outcome is different from that of the pure state. Furthermore, the steerability of entangled mixed states is weaker than that of entangled pure states. Thereby, decoherence can induce the degradation of quantum steering, and the steerability of state is associated with the interaction between quantum systems and reservoirs.

quant-ph↗

arXiv2017

Distributions of a particle's position and their asymptotics in the $q$-deformed totally asymmetric zero range process with site dependent jumping rates

In this paper we study the probability distribution of the position of a tagged particle in the $q$-deformed Totally Asymmetric Zero Range Process ($q$-TAZRP) with site dependent jumping rates. For a finite particle system, it is derived from the transition probability previously obtained by Wang and Waugh. We also provide the probability distribution formula for a tagged particle in the $q$-TAZRP with the so-called step initial condition in which infinitely many particles occupy one single site and all other sites are unoccupied. For the $q$-TAZRP with step initial condition, we provide a Fredholm determinant representation for the probability distribution function of the position of a tagged particle, and moreover we obtain the limiting distribution function as the time goes to infinity. Our asymptotic result for $q$-TAZRP with step initial condition is comparable to the limiting distribution function obtained by Tracy and Widom for the $k$-th leftmost particle in the asymmetric simple exclusion process with step initial condition (Theorem 2 in Commun. Math. Phys. 290, 129--154 (2009)).

math.PR↗

arXiv2017

Entropic uncertainty relations for Markovian and non-Markovian processes under a structured bosonic reservoir

The uncertainty relation is a fundamental limit in quantum mechanics and is of great importance to quantum information processing as it relates to quantum precision measurement. Due to interactions with the surrounding environment, a quantum system will unavoidably suffer from decoherence. Here, we investigate the dynamic behaviors of the entropic uncertainty relation of an atom-cavity interacting system under a bosonic reservoir during the crossover between Markovian and non-Markovian regimes. Specifically, we explore the dynamic behavior of the entropic uncertainty relation for a pair of incompatible observables under the reservoir-induced atomic decay effect both with and without quantum memory. We find that the uncertainty dramatically depends on both the atom-cavity and the cavity-reservoir interactions, as well as the correlation time, $τ$, of the structured reservoir. Furthermore, we verify that the uncertainty is anti-correlated with the purity of the state of the observed qubit-system. We also propose a remarkably simple and efficient way to reduce the uncertainty by utilizing quantum weak measurement reversal. Therefore our work offers a new insight into the uncertainty dynamics for multi-component measurements within an open system, and is thus important for quantum precision measurements.

quant-ph↗

arXiv2017

Memory Visualization for Gated Recurrent Neural Networks in Speech Recognition

Recurrent neural networks (RNNs) have shown clear superiority in sequence modeling, particularly the ones with gated units, such as long short-term memory (LSTM) and gated recurrent unit (GRU). However, the dynamic properties behind the remarkable performance remain unclear in many applications, e.g., automatic speech recognition (ASR). This paper employs visualization techniques to study the behavior of LSTM and GRU when performing speech recognition tasks. Our experiments show some interesting patterns in the gated memory, and some of them have inspired simple yet effective modifications on the network structure. We report two of such modifications: (1) lazy cell update in LSTM, and (2) shortcut connections for residual learning. Both modifications lead to more comprehensible and powerful networks.

cs.LG↗

arXiv2017

A beam monitor using silicon pixel sensors for hadron therapy

We report the design and test results of a beam monitor developed for online monitoring in hadron therapy. The beam monitor uses eight silicon pixel sensors, \textit{Topmetal-${II}^-$}, as the anode array. \textit{Topmetal-${II}^-$} is a charge sensor designed in a CMOS 0.35 $μ$m technology. Each \textit{Topmetal-${II}^-$} sensor has $72\times72$ pixels and the pixel size is $83\times83$ $μ$m$^2$. In our design, the beam passes through the beam monitor without hitting the electrodes, making the beam monitor especially suitable for monitoring heavy ion beams. This design also reduces radiation damage to the beam monitor itself. The beam monitor is tested with a carbon ion beam at the Heavy Ion Research Facility in Lanzhou (HIRFL). Results indicate that the beam monitor can measure position, incidence angle and intensity of the beam with a position resolution better than 20 $μ$m, angular resolution about 0.5$^\circ$ and intensity statistical accuracy better than 2$\%$.

physics.ins-det↗

arXiv2017

Extending the photon energy coverage of an x-ray self-seeding FEL via the reverse taper enhanced harmonic generation technique

In this paper, a simple method is proposed to extend the photon energy range of a soft x-ray self-seeding free-electron laser (FEL). A normal monochromator is first applied to purify the FEL spectrum and provide a coherent seeding signal. This coherent signal then interacts with the electron beam in the following reverse tapered undulator section to generate strong coherent microbunchings while maintain the good quality of the electron beam. After that, the pre-bunched electron beam is sent into the third undulator section which resonates at a target high harmonic of the seed to amplify the coherent radiation at shorter wavelength. Three dimensional simulations have been performed and the results demonstrate that the photon energy gap between 1.5 keV and 4.5 keV of the self-seeding scheme can be fully covered and 100 GW-level peak power can be achieved by using the proposed technique.

physics.acc-ph↗

arXiv2017

A highly pixelated CdZnTe detector based on \textit{Topmetal-${II}^-$} sensor

\textit{Topmetal-${II}^-$} is a low noise CMOS pixel direct charge sensor with a pitch of 83$μm$. CdZnTe is an excellent semiconductor material for radiation detection. The combination of CdZnTe and the sensor makes it possible to build a detector with high spatial resolution. In our experiments, an epoxy adhesive is used as the conductive medium to connect the sensor and Cadmium Zinc Telluride (CdZnTe). The diffusion coefficient and charge efficiency of electrons are measured at a low bias voltage of -2 Volts, and the image of a single alpha is clear with a reasonable spatial resolution. The detector of such structure has the potential to be applied in X-ray imaging systems with a further improvements of the sensor.

physics.ins-det↗