arXiv subjects

Dan Su

Publications and source records attributed to Dan Su.

At least 91 records · Page 5Linked to original sources

arXiv2021

Tune-In: Training Under Negative Environments with Interference for Attention Networks Simulating Cocktail Party Effect

We study the cocktail party problem and propose a novel attention network called Tune-In, abbreviated for training under negative environments with interference. It firstly learns two separate spaces of speaker-knowledge and speech-stimuli based on a shared feature space, where a new block structure is designed as the building block for all spaces, and then cooperatively solves different tasks. Between the two spaces, information is cast towards each other via a novel cross- and dual-attention mechanism, mimicking the bottom-up and top-down processes of a human's cocktail party effect. It turns out that substantially discriminative and generalizable speaker representations can be learnt in severely interfered conditions via our self-supervised training. The experimental results verify this seeming paradox. The learnt speaker embedding has superior discriminative power than a standard speaker verification method; meanwhile, Tune-In achieves remarkably better speech separation performances in terms of SI-SNRi and SDRi consistently in all test modes, and especially at lower memory and computational consumption, than state-of-the-art benchmark systems.

eess.AS↗

arXiv2021

Contrastive Separative Coding for Self-supervised Representation Learning

To extract robust deep representations from long sequential modeling of speech data, we propose a self-supervised learning approach, namely Contrastive Separative Coding (CSC). Our key finding is to learn such representations by separating the target signal from contrastive interfering signals. First, a multi-task separative encoder is built to extract shared separable and discriminative embedding; secondly, we propose a powerful cross-attention mechanism performed over speaker representations across various interfering conditions, allowing the model to focus on and globally aggregate the most critical information to answer the "query" (current bottom-up embedding) while paying less attention to interfering, noisy, or irrelevant parts; lastly, we form a new probabilistic contrastive loss which estimates and maximizes the mutual information between the representations and the global speaker vector. While most prior unsupervised methods have focused on predicting the future, neighboring, or missing samples, we take a different perspective of predicting the interfered samples. Moreover, our contrastive separative loss is free from negative sampling. The experiment demonstrates that our approach can learn useful representations achieving a strong speaker verification performance in adverse conditions.

eess.AS↗

arXiv2021

Replay and Synthetic Speech Detection with Res2net Architecture

Existing approaches for replay and synthetic speech detection still lack generalizability to unseen spoofing attacks. This work proposes to leverage a novel model structure, so-called Res2Net, to improve the anti-spoofing countermeasure's generalizability. Res2Net mainly modifies the ResNet block to enable multiple feature scales. Specifically, it splits the feature maps within one block into multiple channel groups and designs a residual-like connection across different channel groups. Such connection increases the possible receptive fields, resulting in multiple feature scales. This multiple scaling mechanism significantly improves the countermeasure's generalizability to unseen spoofing attacks. It also decreases the model size compared to ResNet-based models. Experimental results show that the Res2Net model consistently outperforms ResNet34 and ResNet50 by a large margin in both physical access (PA) and logical access (LA) of the ASVspoof 2019 corpus. Moreover, integration with the squeeze-and-excitation (SE) block can further enhance performance. For feature engineering, we investigate the generalizability of Res2Net combined with different acoustic features, and observe that the constant-Q transform (CQT) achieves the most promising performance in both PA and LA scenarios. Our best single system outperforms other state-of-the-art single systems in both PA and LA of the ASVspoof 2019 corpus.

eess.AS↗

arXiv2021

VARA-TTS: Non-Autoregressive Text-to-Speech Synthesis based on Very Deep VAE with Residual Attention

This paper proposes VARA-TTS, a non-autoregressive (non-AR) text-to-speech (TTS) model using a very deep Variational Autoencoder (VDVAE) with Residual Attention mechanism, which refines the textual-to-acoustic alignment layer-wisely. Hierarchical latent variables with different temporal resolutions from the VDVAE are used as queries for residual attention module. By leveraging the coarse global alignment from previous attention layer as an extra input, the following attention layer can produce a refined version of alignment. This amortizes the burden of learning the textual-to-acoustic alignment among multiple attention layers and outperforms the use of only a single attention layer in robustness. An utterance-level speaking speed factor is computed by a jointly-trained speaking speed predictor, which takes the mean-pooled latent variables of the coarsest layer as input, to determine number of acoustic frames at inference. Experimental results show that VARA-TTS achieves slightly inferior speech quality to an AR counterpart Tacotron 2 but an order-of-magnitude speed-up at inference; and outperforms an analogous non-AR model, BVAE-TTS, in terms of speech quality.

cs.SD↗

arXiv2021

Multi-hop Question Generation with Graph Convolutional Network

Multi-hop Question Generation (QG) aims to generate answer-related questions by aggregating and reasoning over multiple scattered evidence from different paragraphs. It is a more challenging yet under-explored task compared to conventional single-hop QG, where the questions are generated from the sentence containing the answer or nearby sentences in the same paragraph without complex reasoning. To address the additional challenges in multi-hop QG, we propose Multi-Hop Encoding Fusion Network for Question Generation (MulQG), which does context encoding in multiple hops with Graph Convolutional Network and encoding fusion via an Encoder Reasoning Gate. To the best of our knowledge, we are the first to tackle the challenge of multi-hop reasoning over paragraphs without any sentence-level information. Empirical results on HotpotQA dataset demonstrate the effectiveness of our method, in comparison with baselines on automatic evaluation metrics. Moreover, from the human evaluation, our proposed model is able to generate fluent questions with high completeness and outperforms the strongest baseline by 20.8% in the multi-hop evaluation. The code is publicly available at https://github.com/HLTCHKUST/MulQG}{https://github.com/HLTCHKUST/MulQG .

cs.CL↗

arXiv2021

Effective Low-Cost Time-Domain Audio Separation Using Globally Attentive Locally Recurrent Networks

Recent research on the time-domain audio separation networks (TasNets) has brought great success to speech separation. Nevertheless, conventional TasNets struggle to satisfy the memory and latency constraints in industrial applications. In this regard, we design a low-cost high-performance architecture, namely, globally attentive locally recurrent (GALR) network. Alike the dual-path RNN (DPRNN), we first split a feature sequence into 2D segments and then process the sequence along both the intra- and inter-segment dimensions. Our main innovation lies in that, on top of features recurrently processed along the inter-segment dimensions, GALR applies a self-attention mechanism to the sequence along the inter-segment dimension, which aggregates context-aware information and also enables parallelization. Our experiments suggest that GALR is a notably more effective network than the prior work. On one hand, with only 1.5M parameters, it has achieved comparable separation performance at a much lower cost with 36.1% less runtime memory and 49.4% fewer computational operations, relative to the DPRNN. On the other hand, in a comparable model size with DPRNN, GALR has consistently outperformed DPRNN in three datasets, in particular, with a substantial margin of 2.4dB absolute improvement of SI-SNRi in the benchmark WSJ0-2mix task.

eess.AS↗

arXiv2020

CAiRE-COVID: A Question Answering and Query-focused Multi-Document Summarization System for COVID-19 Scholarly Information Management

We present CAiRE-COVID, a real-time question answering (QA) and multi-document summarization system, which won one of the 10 tasks in the Kaggle COVID-19 Open Research Dataset Challenge, judged by medical experts. Our system aims to tackle the recent challenge of mining the numerous scientific articles being published on COVID-19 by answering high priority questions from the community and summarizing salient question-related information. It combines information extraction with state-of-the-art QA and query-focused multi-document summarization techniques, selecting and highlighting evidence snippets from existing literature given a query. We also propose query-focused abstractive and extractive multi-document summarization methods, to provide more relevant information related to the question. We further conduct quantitative experiments that show consistent improvements on various metrics for each module. We have launched our website CAiRE-COVID for broader use by the medical community, and have open-sourced the code for our system, to bootstrap further study by other researches.

cs.CL↗

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

Audio-visual Multi-channel Recognition of Overlapped Speech

Automatic speech recognition (ASR) of overlapped speech remains a highly challenging task to date. To this end, multi-channel microphone array data are widely used in state-of-the-art ASR systems. Motivated by the invariance of visual modality to acoustic signal corruption, this paper presents an audio-visual multi-channel overlapped speech recognition system featuring tightly integrated separation front-end and recognition back-end. A series of audio-visual multi-channel speech separation front-end components based on \textit{TF masking}, \textit{filter\&sum} and \textit{mask-based MVDR} beamforming approaches were developed. To reduce the error cost mismatch between the separation and recognition components, they were jointly fine-tuned using the connectionist temporal classification (CTC) loss function, or a multi-task criterion interpolation with scale-invariant signal to noise ratio (Si-SNR) error cost. Experiments suggest that the proposed multi-channel AVSR system outperforms the baseline audio-only ASR system by up to 6.81\% (26.83\% relative) and 22.22\% (56.87\% relative) absolute word error rate (WER) reduction on overlapped speech constructed using either simulation or replaying of the lipreading sentence 2 (LRS2) dataset respectively.

eess.AS↗

arXiv2020

"Are you home alone?" "Yes" Disclosing Security and Privacy Vulnerabilities in Alexa Skills

The home voice assistants such as Amazon Alexa have become increasingly popular due to many interesting voice-activated services provided through special applications called skills. These skills, though useful, have also introduced new security and privacy challenges. Prior work has verified that Alexa is vulnerable to multiple types of voice attacks, but the security and privacy risk of using skills has not been fully investigated. In this work, we study an adversary model that covers three severe privacy-related vulnerabilities, namely,over-privileged resource access, hidden code-manipulation and hidden content-manipulation. By exploiting these vulnerabilities, malicious skills can not only bypass the security tests in the vetting process, but also surreptitiously change their original functions in an attempt to steal users' personal information. What makes the situation even worse is that the attacks can be extended from virtual networks to the physical world. We systematically study the security issues from the feasibility and implementation of the attacks to the design of countermeasures. We also made a comprehensive survey study of 33,744 skills in Alex Skills Store.

cs.CR↗

arXiv2020

Dimsum @LaySumm 20: BART-based Approach for Scientific Document Summarization

Lay summarization aims to generate lay summaries of scientific papers automatically. It is an essential task that can increase the relevance of science for all of society. In this paper, we build a lay summary generation system based on the BART model. We leverage sentence labels as extra supervision signals to improve the performance of lay summarization. In the CL-LaySumm 2020 shared task, our model achieves 46.00\% Rouge1-F1 score.

cs.CL↗

arXiv2020

Investigating Robustness of Adversarial Samples Detection for Automatic Speaker Verification

Recently adversarial attacks on automatic speaker verification (ASV) systems attracted widespread attention as they pose severe threats to ASV systems. However, methods to defend against such attacks are limited. Existing approaches mainly focus on retraining ASV systems with adversarial data augmentation. Also, countermeasure robustness against different attack settings are insufficiently investigated. Orthogonal to prior approaches, this work proposes to defend ASV systems against adversarial attacks with a separate detection network, rather than augmenting adversarial data into ASV training. A VGG-like binary classification detector is introduced and demonstrated to be effective on detecting adversarial samples. To investigate detector robustness in a realistic defense scenario where unseen attack settings may exist, we analyze various kinds of unseen attack settings' impact and observe that the detector is robust (6.27\% EER_{det} degradation in the worst case) against unseen substitute ASV systems, but it has weak robustness (50.37\% EER_{det} degradation in the worst case) against unseen perturbation methods. The weak robustness against unseen perturbation methods shows a direction for developing stronger countermeasures.

eess.AS↗

arXiv2020

Distortionless Multi-Channel Target Speech Enhancement for Overlapped Speech Recognition

Speech enhancement techniques based on deep learning have brought significant improvement on speech quality and intelligibility. Nevertheless, a large gain in speech quality measured by objective metrics, such as perceptual evaluation of speech quality (PESQ), does not necessarily lead to improved speech recognition performance due to speech distortion in the enhancement stage. In this paper, a multi-channel dilated convolutional network based frequency domain modeling is presented to enhance target speaker in the far-field, noisy and multi-talker conditions. We study three approaches towards distortionless waveforms for overlapped speech recognition: estimating complex ideal ratio mask with an infinite range, incorporating the fbank loss in a multi-objective learning and finetuning the enhancement model by an acoustic model. Experimental results proved the effectiveness of all three approaches on reducing speech distortions and improving recognition accuracy. Particularly, the jointly tuned enhancement model works very well with other standalone acoustic model on real test data.

eess.AS↗

arXiv2020

Speaker Independent and Multilingual/Mixlingual Speech-Driven Talking Head Generation Using Phonetic Posteriorgrams

Generating 3D speech-driven talking head has received more and more attention in recent years. Recent approaches mainly have following limitations: 1) most speaker-independent methods need handcrafted features that are time-consuming to design or unreliable; 2) there is no convincing method to support multilingual or mixlingual speech as input. In this work, we propose a novel approach using phonetic posteriorgrams (PPG). In this way, our method doesn't need hand-crafted features and is more robust to noise compared to recent approaches. Furthermore, our method can support multilingual speech as input by building a universal phoneme space. As far as we know, our model is the first to support multilingual/mixlingual speech as input with convincing results. Objective and subjective experiments have shown that our model can generate high quality animations given speech from unseen languages or speakers and be robust to noise.

eess.AS↗

arXiv2020

Speech-XLNet: Unsupervised Acoustic Model Pretraining For Self-Attention Networks

Self-attention network (SAN) can benefit significantly from the bi-directional representation learning through unsupervised pretraining paradigms such as BERT and XLNet. In this paper, we present an XLNet-like pretraining scheme "Speech-XLNet" for unsupervised acoustic model pretraining to learn speech representations with SAN. The pretrained SAN is finetuned under the hybrid SAN/HMM framework. We conjecture that by shuffling the speech frame orders, the permutation in Speech-XLNet serves as a strong regularizer to encourage the SAN to make inferences by focusing on global structures through its attention weights. In addition, Speech-XLNet also allows the model to explore the bi-directional contexts for effective speech representation learning. Experiments on TIMIT and WSJ demonstrate that Speech-XLNet greatly improves the SAN/HMM performance in terms of both convergence speed and recognition accuracy compared to the one trained from randomly initialized weights. Our best systems achieve a relative improvement of 11.9% and 8.3% on the TIMIT and WSJ tasks respectively. In particular, the best system achieves a phone error rate (PER) of 13.3% on the TIMIT test set, which to our best knowledge, is the lowest PER obtained from a single system.

cs.CL↗

arXiv2020

End-to-End Multi-Look Keyword Spotting

The performance of keyword spotting (KWS), measured in false alarms and false rejects, degrades significantly under the far field and noisy conditions. In this paper, we propose a multi-look neural network modeling for speech enhancement which simultaneously steers to listen to multiple sampled look directions. The multi-look enhancement is then jointly trained with KWS to form an end-to-end KWS model which integrates the enhanced signals from multiple look directions and leverages an attention mechanism to dynamically tune the model's attention to the reliable sources. We demonstrate, on our large noisy and far-field evaluation sets, that the proposed approach significantly improves the KWS performance against the baseline KWS system and a recent beamformer based multi-beam KWS system.

eess.AS↗

arXiv2020

Transferring Source Style in Non-Parallel Voice Conversion

Voice conversion (VC) techniques aim to modify speaker identity of an utterance while preserving the underlying linguistic information. Most VC approaches ignore modeling of the speaking style (e.g. emotion and emphasis), which may contain the factors intentionally added by the speaker and should be retained during conversion. This study proposes a sequence-to-sequence based non-parallel VC approach, which has the capability of transferring the speaking style from the source speech to the converted speech by explicitly modeling. Objective evaluation and subjective listening tests show superiority of the proposed VC approach in terms of speech naturalness and speaker similarity of the converted speech. Experiments are also conducted to show the source-style transferability of the proposed approach.

eess.AS↗

arXiv2020

Enhancing End-to-End Multi-channel Speech Separation via Spatial Feature Learning

Hand-crafted spatial features (e.g., inter-channel phase difference, IPD) play a fundamental role in recent deep learning based multi-channel speech separation (MCSS) methods. However, these manually designed spatial features are hard to incorporate into the end-to-end optimized MCSS framework. In this work, we propose an integrated architecture for learning spatial features directly from the multi-channel speech waveforms within an end-to-end speech separation framework. In this architecture, time-domain filters spanning signal channels are trained to perform adaptive spatial filtering. These filters are implemented by a 2d convolution (conv2d) layer and their parameters are optimized using a speech separation objective function in a purely data-driven fashion. Furthermore, inspired by the IPD formulation, we design a conv2d kernel to compute the inter-channel convolution differences (ICDs), which are expected to provide the spatial cues that help to distinguish the directional sources. Evaluation results on simulated multi-channel reverberant WSJ0 2-mix dataset demonstrate that our proposed ICD based MCSS model improves the overall signal-to-distortion ratio by 10.4% over the IPD based MCSS model.

eess.AS↗