arXiv subjects

Chunlei Zhang

Publications and source records attributed to Chunlei Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2021

Towards Robust Speaker Verification with Target Speaker Enhancement

This paper proposes the target speaker enhancement based speaker verification network (TASE-SVNet), an all neural model that couples target speaker enhancement and speaker embedding extraction for robust speaker verification (SV). Specifically, an enrollment speaker conditioned speech enhancement module is employed as the front-end for extracting target speaker from its mixture with interfering speakers and environmental noises. Compared with the conventional target speaker enhancement models, nontarget speaker/interference suppression should draw additional attention for SV. Therefore, an effective nontarget speaker sampling strategy is explored. To improve speaker embedding extraction with a light-weighted model, a teacher-student (T/S) training is proposed to distill speaker discriminative information from large models to small models. Iterative inference is investigated to address the noisy speaker enrollment problem. We evaluate the proposed method on two SV tasks, i.e., one heavily overlapped speech and the other one with comprehensive noise types in vehicle environments. Experiments show significant and consistent improvements in Equal Error Rate (EER) over the state-of-the-art baselines.

eess.AS↗

arXiv2021

Improving RNN Transducer With Target Speaker Extraction and Neural Uncertainty Estimation

Target-speaker speech recognition aims to recognize target-speaker speech from noisy environments with background noise and interfering speakers. This work presents a joint framework that combines time-domain target-speaker speech extraction and Recurrent Neural Network Transducer (RNN-T). To stabilize the joint-training, we propose a multi-stage training strategy that pre-trains and fine-tunes each module in the system before joint-training. Meanwhile, speaker identity and speech enhancement uncertainty measures are proposed to compensate for residual noise and artifacts from the target speech extraction module. Compared to a recognizer fine-tuned with a target speech extraction model, our experiments show that adding the neural uncertainty module significantly reduces 17% relative Character Error Rate (CER) on multi-speaker signals with background noise. The multi-condition experiments indicate that our method can achieve 9% relative performance gain in the noisy condition while maintaining the performance in the clean condition.

cs.SD↗

arXiv2021

Self-supervised Text-independent Speaker Verification using Prototypical Momentum Contrastive Learning

In this study, we investigate self-supervised representation learning for speaker verification (SV). First, we examine a simple contrastive learning approach (SimCLR) with a momentum contrastive (MoCo) learning framework, where the MoCo speaker embedding system utilizes a queue to maintain a large set of negative examples. We show that better speaker embeddings can be learned by momentum contrastive learning. Next, alternative augmentation strategies are explored to normalize extrinsic speaker variabilities of two random segments from the same speech utterance. Specifically, augmentation in the waveform largely improves the speaker representations for SV tasks. The proposed MoCo speaker embedding is further improved when a prototypical memory bank is introduced, which encourages the speaker embeddings to be closer to their assigned prototypes with an intermediate clustering step. In addition, we generalize the self-supervised framework to a semi-supervised scenario where only a small portion of the data is labeled. Comprehensive experiments on the Voxceleb dataset demonstrate that our proposed self-supervised approach achieves competitive performance compared with existing techniques, and can approach fully supervised results with partially labeled data.

eess.AS↗

arXiv2020

Phonetic Posteriorgrams based Many-to-Many Singing Voice Conversion via Adversarial Training

This paper describes an end-to-end adversarial singing voice conversion (EA-SVC) approach. It can directly generate arbitrary singing waveform by given phonetic posteriorgram (PPG) representing content, F0 representing pitch, and speaker embedding representing timbre, respectively. Proposed system is composed of three modules: generator $G$, the audio generation discriminator $D_{A}$, and the feature disentanglement discriminator $D_F$. The generator $G$ encodes the features in parallel and inversely transforms them into the target waveform. In order to make timbre conversion more stable and controllable, speaker embedding is further decomposed to the weighted sum of a group of trainable vectors representing different timbre clusters. Further, to realize more robust and accurate singing conversion, disentanglement discriminator $D_F$ is proposed to remove pitch and timbre related information that remains in the encoded PPG. Finally, a two-stage training is conducted to keep a stable and effective adversarial training process. Subjective evaluation results demonstrate the effectiveness of our proposed methods. Proposed system outperforms conventional cascade approach and the WaveNet based end-to-end approach in terms of both singing quality and singer similarity. Further objective analysis reveals that the model trained with the proposed two-stage training strategy can produce a smoother and sharper formant which leads to higher audio quality.

cs.SD↗

arXiv2020

DurIAN-SC: Duration Informed Attention Network based Singing Voice Conversion System

Singing voice conversion is converting the timbre in the source singing to the target speaker's voice while keeping singing content the same. However, singing data for target speaker is much more difficult to collect compared with normal speech data.In this paper, we introduce a singing voice conversion algorithm that is capable of generating high quality target speaker's singing using only his/her normal speech data. First, we manage to integrate the training and conversion process of speech and singing into one framework by unifying the features used in standard speech synthesis system and singing synthesis system. In this way, normal speech data can also contribute to singing voice conversion training, making the singing voice conversion system more robust especially when the singing database is small.Moreover, in order to achieve one-shot singing voice conversion, a speaker embedding module is developed using both speech and singing data, which provides target speaker identify information during conversion. Experiments indicate proposed sing conversion system can convert source singing to target speaker's high-quality singing with only 20 seconds of target speaker's enrollment speech data.

eess.AS↗

arXiv2019

Minimum Bayes Risk Training of RNN-Transducer for End-to-End Speech Recognition

In this work, we propose minimum Bayes risk (MBR) training of RNN-Transducer (RNN-T) for end-to-end speech recognition. Specifically, initialized with a RNN-T trained model, MBR training is conducted via minimizing the expected edit distance between the reference label sequence and on-the-fly generated N-best hypothesis. We also introduce a heuristic to incorporate an external neural network language model (NNLM) in RNN-T beam search decoding and explore MBR training with the external NNLM. Experimental results demonstrate an MBR trained model outperforms a RNN-T trained model substantially and further improvements can be achieved if trained with an external NNLM. Our best MBR trained system achieves absolute character error rate (CER) reductions of 1.2% and 0.5% on read and spontaneous Mandarin speech respectively over a strong convolution and transformer based RNN-T baseline trained on ~21,000 hours of speech.

cs.CL↗

arXiv2019

The Medium Energy (ME) X-ray telescope onboard the Insight-HXMT astronomy satellite

The Medium Energy X-ray telescope (ME) is one of the three main telescopes on board the Insight Hard X-ray Modulation Telescope (Insight-HXMT) astronomy satellite. ME contains 1728 pixels of Si-PIN detectors sensitive in 5-30 keV with a total geometrical area of 952 cm2. Application Specific Integrated Circuit (ASIC) chips, VA32TA6, is used to achieve low power consumption and low readout noise. The collimators define three kinds of field of views (FOVs) for the telescope, 1°{\times}4°, 4°{\times}4°, and blocked ones. Combination of such FOVs can be used to estimate the in-orbit X-ray and particle background components. The energy resolution of ME is ~3 keV at 17.8 keV (FWHM) and the time resolution is 255 μs. In this paper, we introduce the design and performance of ME.

astro-ph.IM↗

arXiv2019

Spin current pumped by resonant skyrmion

Spin pumping is a widely recognized method to generate the spin current in the spintronics, which is acknowledged as a fundamentally dynamic process equivalent to the spin-transfer torque. In this work, we theoretically verify that the oscillating spin current can be pumped from the microwave-motivated breathing skyrmion. The skyrmion spin pumping can be excited by a relatively low frequency compared with the ferromagnetic resonance (FMR) and the current density is larger than the ordinary FMR spin pumping. Based on the skyrmion spin pumping, we build a high reading-speed racetrack memory model whose reading speed is an order of magnitude higher than the SOT (spin-orbit torque) /STT (spin-transfer torque) skyrmion racetrack. Our work explored the spin pumping phenomenon in the skyrmion, and it may contribute to the applications of the skyrmion-based device.

cond-mat.mes-hall↗

arXiv2019

Gilbert damping of [Co/Pd]n/Py multilayer thin films

Understanding the Gilbert damping in exchange-coupled multilayer materials is particularly important to develop future fast switching spintronics devices. Here, we report an experimental investigation of temperature-dependent Gilbert damping in [Co/Pd]n/Py multilayer films of varying the number of Co/Pd repetitions by ferromagnetic resonance. The results demonstrate that three independent contributions to the Gilbert damping are identified, namely the intrinsic Gilbert damping, the inhomogeneous linewidth broadening and the two-magnon scattering contribution. Of particular interest, the two-magnon scattering intensity increases as the enlargement of number repetitions of Co/Pd due to the larger pinning effect at the interface between Py and the Co/Pd layers. The Gilbert damping increases monotonically as the temperature decreases from 300K to 50K. Our findings open the door to comprehend the physical origin of the Gilbert damping in ultrathin exchange-coupled multilayer films.

physics.app-ph↗

arXiv2019

Insight-HXMT observations of Swift J0243.6+6124 during its 2017-2018 outburst

The recently discovered neutron star transient Swift J0243.6+6124 has been monitored by {\it the Hard X-ray Modulation Telescope} ({\it Insight-\rm HXMT). Based on the obtained data, we investigate the broadband spectrum of the source throughout the outburst. We estimate the broadband flux of the source and search for possible cyclotron line in the broadband spectrum. No evidence of line-like features is, however, found up to $\rm 150~keV$. In the absence of any cyclotron line in its energy spectrum, we estimate the magnetic field of the source based on the observed spin evolution of the neutron star by applying two accretion torque models. In both cases, we get consistent results with $B\rm \sim 10^{13}~G$, $D\rm \sim 6~kpc$ and peak luminosity of $\rm >10^{39}~erg~s^{-1}$ which makes the source the first Galactic ultraluminous X-ray source hosting a neutron star.

astro-ph.HE↗

arXiv2019

I4U Submission to NIST SRE 2018: Leveraging from a Decade of Shared Experiences

The I4U consortium was established to facilitate a joint entry to NIST speaker recognition evaluations (SRE). The latest edition of such joint submission was in SRE 2018, in which the I4U submission was among the best-performing systems. SRE'18 also marks the 10-year anniversary of I4U consortium into NIST SRE series of evaluation. The primary objective of the current paper is to summarize the results and lessons learned based on the twelve sub-systems and their fusion submitted to SRE'18. It is also our intention to present a shared view on the advancements, progresses, and major paradigm shifts that we have witnessed as an SRE participant in the past decade from SRE'08 to SRE'18. In this regard, we have seen, among others, a paradigm shift from supervector representation to deep speaker embedding, and a switch of research challenge from channel compensation to domain adaptation.

eess.AS↗

arXiv2019

Current-induced motion of twisted skyrmions

Twisted skyrmions, whose helicity angles are different from that of Bloch skyrmions and Néel skyrmions, have already been demonstrated in experiments recently. In this work, we first contrast the magnetic structure and origin of the twisted skyrmion with other three types of skyrmion including Bloch skyrmion, Néel skyrmion and antiskyrmion. Following, we investigate the dynamics of twisted skyrmions driven by the spin transfer toque (STT) and the spin Hall effect (SHE) by using micromagnetic simulations. It is found that the spin Hall angle of the twisted skyrmion is related to the dissipative force tensor and the Gilbert damping both for the motions induced by the STT and the SHE, especially for the SHE induced motion, the skyrmion Hall angle depends substantially on the skyrmion helicity. At last, we demonstrate that the trajectory of the twisted skyrmion can be controlled in a two dimensional plane with a Gilbert damping gradient. Our results provide the understanding of current-induced motion of twisted skyrmions, which may contribute to the applications of skyrmion-based racetrack memories.

physics.app-ph↗

arXiv2019

Directional Spin-wave Propagation in the Skyrmion Chain

The Dzyaloshinskii-Moriya interaction (DMI), favoring a chiral spin structure like the skyrmion, gives rise to the nonreciprocal propagation of spin waves. We investigate the propagation of spin waves in a nanostripe with the presence of a skyrmion chain by using micromagnetic simulations. Through applying a microwave locally, it is found that when the interval between skyrmions is large enough, the spin waves can be separated to the counter direction according to different frequencies. While for the tightly arranged skyrmions, the skyrmion chain with strong interactions between skyrmions becomes a channel for spin waves, which is around the frequency of skyrmion breathing and exhibit a characteristic of directional propagation. This work opens a vista for skyrmion-based spin wave devices.

cond-mat.mes-hall↗

arXiv2018

Metastable magnetic bubble in [Co/Pd]4/Py multilayers

Magnetic bubbles are topologically spin textures that offering the interesting physics and great promise for next-generation information storage technologies. The main obstacles so far are that magnetic bubbles are generated with no field stimuli in new material systems at room temperature. Here, we report the observation of individual magnetic bubbles and its high frequency measurement at room temperature in an exchange-coupled [Co/Pd]4/Py multilayers. We demonstrate that the emergence of magnetic bubbles at remanence can be tuned by the in-plane tilted magnetic field (roughly 3°) along the film plane at room temperature. High frequency results indicate that the presence of magnetic bubbles leads to broadening of the magnetic permeability spectrum lines (due to the non-uniformity of the magnetic moments). Our findings open the door to the bubble-based spintronics at room temperature in exchange-coupled magnetic multilayers.

physics.app-ph↗

arXiv2018

Privacy-preserving Decentralized Optimization via Decomposition

This paper considers the problem of privacy-preservation in decentralized optimization, in which $N$ agents cooperatively minimize a global objective function that is the sum of $N$ local objective functions. We assume that each local objective function is private and only known to an individual agent. To cooperatively solve the problem, most existing decentralized optimization approaches require participating agents to exchange and disclose estimates to neighboring agents. However, this results in leakage of private information about local objective functions, which is undesirable when adversaries exist and try to steal information from participating agents. To address this issue, we propose a privacy-preserving decentralized optimization approach based on proximal Jacobian ADMM via function decomposition. Numerical simulations confirm the effectiveness of the proposed approach.

math.OC↗

arXiv2018

Distributed Event Localization via Alternating Direction Method of Multipliers

This paper addresses the problem of distributed event localization using noisy range measurements with respect to sensors with known positions. Event localization is fundamental in many wireless sensor network applications such as homeland security, law enforcement, and environmental studies. However, most existing distributed algorithms require the target event to be within the convex hull of the deployed sensors. Based on the alternating direction method of multipliers (ADMM), we propose two scalable distributed algorithms named GS-ADMM and J-ADMM which do not require the target event to be within the convex hull of the deployed sensors. More specifically, the two algorithms can be implemented in a scenario in which the entire sensor network is divided into several clusters with cluster heads collecting measurements within each cluster and exchanging intermediate computation information to achieve localization consistency (consensus) across all clusters. This scenario is important in many applications such as homeland security and law enforcement. Simulation results confirm effectiveness of the proposed algorithms.

cs.DC↗

arXiv2018

ADMM Based Privacy-preserving Decentralized Optimization

Privacy preservation is addressed for decentralized optimization, where $N$ agents cooperatively minimize the sum of $N$ convex functions private to these individual agents. In most existing decentralized optimization approaches, participating agents exchange and disclose states explicitly, which may not be desirable when the states contain sensitive information of individual agents. The problem is more acute when adversaries exist which try to steal information from other participating agents. To address this issue, we propose a privacy-preserving decentralized optimization approach based on ADMM and partially homomorphic cryptography. To our knowledge, this is the first time that cryptographic techniques are incorporated in a fully decentralized setting to enable privacy preservation in decentralized optimization in the absence of any third party or aggregator. To facilitate the incorporation of encryption in a fully decentralized manner, we introduce a new ADMM which allows time-varying penalty matrices and rigorously prove that it has a convergence rate of $O(1/t)$. Numerical and experimental results confirm the effectiveness and low computational complexity of the proposed approach.

math.OC↗

arXiv2017

UTD-CRSS Submission for MGB-3 Arabic Dialect Identification: Front-end and Back-end Advancements on Broadcast Speech

This study presents systems submitted by the University of Texas at Dallas, Center for Robust Speech Systems (UTD-CRSS) to the MGB-3 Arabic Dialect Identification (ADI) subtask. This task is defined to discriminate between five dialects of Arabic, including Egyptian, Gulf, Levantine, North African, and Modern Standard Arabic. We develop multiple single systems with different front-end representations and back-end classifiers. At the front-end level, feature extraction methods such as Mel-frequency cepstral coefficients (MFCCs) and two types of bottleneck features (BNF) are studied for an i-Vector framework. As for the back-end level, Gaussian back-end (GB), and Generative Adversarial Networks (GANs) classifiers are applied alternately. The best submission (contrastive) is achieved for the ADI subtask with an accuracy of 76.94% by augmenting the randomly chosen part of the development dataset. Further, with a post evaluation correction in the submitted system, final accuracy is increased to 79.76%, which represents the best performance achieved so far for the challenge on the test dataset.

eess.AS↗