yuyq96/icassp2019

Papers in ICASSP 2019

ICASSP 2019

Speech Language Processing

Lecture 1: End-to-end Speech Recognition I: General Topics

• Bo Li, Yu Zhang, Tara Sainath, Yonghui Wu, William Chan. Bytes Are All You Need: End-to-end Multilingual Speech Recognition and Synthesis with Bytes. [ICASSP 2019] [CoRR2018]

  • ASR TTS Multilingual
  • Use Unicode bytes instead of characters, sub-words or words as the unit of text representation.
  • ASR: outperform grapheme models in both multilingual and monolingual models.
  • TTS: match the performance of monolingual grapheme models.
  • Small vocabulary size (256 for UTF-8) helps to build compact models.

• Shuo-Yiin Chang, Rohit Prabhavalkar, Yanzhang He, Tara N. Sainath, Gabor Simko. Joint Endpointing and Decoding with End-to-end Models. [ICASSP 2019]

• Changhao Shan, Chao Weng, Guangsen Wang, Dan Su, Min Luo, Dong Yu, Lei Xie. Component Fusion: Learning Replaceable Language Model Component for End-to-end Speech Recognition System. [ICASSP 2019]

  • ASR LM Fushion Component Fushion
  • Use Component Fusion to incorporate externally trained LM into an attention-based ASR system.
  • Make use of the large amount of text corpora.
  • Concatenate the hidden states of ASR model and LM (unlike Cold Fushion: first, train the ASR model jointly with a fixed LM pre-trained on transcript text, then replace the LM with another one which is trained on a larger text corpora or out-of-domain text during decoding).
  • Achieve better performance in both in-domain and out-of-domain senary.

• Stefan Braun, Shih-Chii Liu. Parameter Uncertainty for End-to-end Speech Recognition. [ICASSP 2019]

• Shane Settle, Kartik Audhkhasi, Karen Livescu, Michael Picheny. Acoustically Grounded Word Embeddings for Improved Acoustics-to-Word Speech Recognition. [ICASSP 2019] [CoRR 2019]

• Murali Karthick Baskar, Lukás Burget, Shinji Watanabe, Martin Karafiát, Takaaki Hori, Jan Honza Cernocký. Promising Accurate Prefix Boosting for sequence-to-sequence ASR. [ICASSP 2019] [CoRR 2018]

Lecture 2: End-to-end Speech Recognition II: New Models

Lecture 6: Systems for Speaker Recognition and Identification

Poster 2: Speaker Verification and Identification I

• Mahesh Kumar Nandwana, Mitchell McLaren, Luciana Ferrer, Diego Castan, Aaron Lawson. Analysis and Mitigation of Vocal Effort Variations in Speaker Recognition. [ICASSP 2019]

• Johan Rohdin, Themos Stafylakis, Anna Silnova, Hossein Zeinali, Lukáš Burget, Oldřich PlchotSpeaker. Speaker Verification Using End-to-End Adversarial Language Adaptation. [ICASSP 2019]

  • TI-SV SRE Domain Adaption GAN Maximum Mean Discrepancy (MMD) Wasserstein Distance
  • Generator: minimize Wasserstein distance and classification loss.
  • Discriminator: maximize Wasserstein distance (and minimize gradient penalty loss for 1-Lipschitz constraint).

• Buddhi Wickramasinghe, Eliathamby Ambikairajah, Julien Epps, Vidhyasaharan Sethu, Haizhou Li. Auditory Inspired Spatial Differentiation for Replay Spoofing Attack Detection. [ICASSP 2019]

  • TI-SV ASVspoof

• Roberto Font. A Denoising Autoencoder for Speaker Recognition. Results on the MCE 2018 Challenge. [ICASSP 2019]

• Shuai Wang, Yexin Yang, Tianzhe Wang, Yanmin Qian, Kai Yu. Knowledge Distillation for Small Foot-print Deep Speaker Embedding. [ICASSP 2019]

  • TI-SV VoxCeleb Knowledge Distillation Teacher-Student
  • Optimize in label level (posterior distributions, by KL-divergence) or embedding level (by Euclidean or cosine distance).

• Liang He, Xianhong Chen, Can Xu, Jia Liu. Multi-objective Optimization Training of PLDA for Speaker Verification. [ICASSP 2019]

• Ondřej Novotný, Oldřich Plchot, Ondřej Glembek, Lukáš Burget, Pavel Matějka. Discriminatively Re-trained i-vector Extractor for Speaker Recognition. [ICASSP 2019] [CoRR 2018]

• Lantian Li, Zhiyuan Tang, Ying Shi, Dong Wang. Gaussian-constrained Training for Speaker Verification. [ICASSP 2019] [CoRR 2018]

• Gautam Bhattacharya, Jahangir Alam, Patrick Kenny. Adapting End-to-End Neural Speaker Verification to New Languages and Recording Conditions with Adversarial Training. [ICASSP 2019] [CoRR 2018]

• Ya-Qi Yu, Lei Fan, Wu-Jun Li. Ensemble Additive Margin Softmax for Speaker Verification. [ICASSP 2019]

  • TI-SV VoxCeleb Additive Margin Softmax Hilbert-Schmidt Independence Criterion

Poster 4: Speaker Verification and Identification II

• Fei Zhao, Hao Li, Xueliang Zhang. A Robust Text-independent Speaker Verification Method Based on Speech Separation and Deep Speaker. [ICASSP 2019]

• Gajan Suthokumar, Kaavya Sriskandaraja, Vidhyasaharan Sethu, Chamith Wijenayake, Eliathamby Ambikairajah. Phoneme Specific Modelling and Scoring Techniques for Anti Spoofing System. [ICASSP 2019]

• Phani Sankar Nidadavolu, Vicente Iglesias, Jesús Villalba, Najim Dehak. Investigation on Neural Bandwidth Extension of Telephone Speech for Improved Speaker Recognition. [ICASSP 2019]

• Yun Tang, Guohong Ding, Jing Huang, Xiaodong He, Bowen Zhou. Deep Speaker Embedding Learning with Multi-level Pooling for Text-independent Speaker Verification. [ICASSP 2019] [CoRR 2018]

• Sandro Cumani, Pietro Laface. Tied Normal Variance–Mean Mixtures for Linear Score Calibration. [ICASSP 2019]

• Insoo Kim, Kyuhong Kim, Jiwhan Kim, Changkyu Choi. Deep Speaker Representation Using Orthogonal Decomposition and Recombination for Speaker Verification. [ICASSP 2019]

• Yichi Zhang, Meng Yu, Na Li, Chengzhu Yu, Jia Cui, Dong Yu. Seq2Seq Attentional Siamese Neural Networks for Text-dependent Speaker Verification. [ICASSP 2019]

  • TD-SV Seq2seq Attention Siamese

• Tharshini Gunendradasan, Saad Irtza, Eliathamby Ambikairajah, Julien Epps. Transmission Line Cochlear Model Based AM-FM Features for Replay Attack Detection. [ICASSP 2019]

• Hossein Zeinali, Luka Burget, Johan Rohdin, Themos Stafylakis, Jan Honza Cernocky. How to Improve Your Speaker Embeddings Extractor in Generic Toolkits. [ICASSP 2019] [CoRR 2018]

• Tomi Kinnunen, Rosa González Hautamäki, Ville Vestman, Md Sahidullah. Can We Use Speaker Recognition Technology to Attack Itself? Enhancing Mimicry Attacks Using Automatic Target Speaker Selection. [ICASSP 2019] [CoRR 2018]

Poster 6: Features and Robustness for Speaker Identification

Poster 8: Features and Learning for Speaker Identification and Diarization

Lecture 8: Speech Synthesis I

• Jean-Marc Valin, Jan Skoglund. LPCNET: Improving Neural Speech Synthesis through Linear Prediction. [ICASSP 2019] [[CoRR 2018](LPCNET: Improving Neural Speech Synthesis through Linear Prediction)]

  • TTS Spectrogram-to-waveform LPCNet
  • Use linear predictor to simplify the prediction, so that the network directly predict the excitation (residual).

• Jungbae Park, Kijong Han, Yuneui Jeong, Sang Wan Lee. Phonemic-level Duration Control Using Attention Alignment for Natural Speech Synthesis. [ICASSP 2019]

• Wei-Ning Hsu, Yu Zhang, Ron J. Weiss, Yu-An Chung, Yuxuan Wang, Yonghui Wu, James Glass. Disentangling Correlated Speaker and Noise for Speech Synthesis via Data Augmentation and Adversarial Factorization. [ICASSP 2019]

• Kyle Kastner, João Felipe Santos, Yoshua Bengio, Aaron Courville. Representation Mixing for TTS Synthesis. [ICASSP 2019] [CoRR 2018]

  • TTS Text-to-spectrogram Representation Mixing
  • Single word can have different pronunciation.
  • Combine grapheme and phoneme inputs in a single encoder flexibly (with mask).

• Younggun Lee, Taesu Kim. Robust and Fine-grained Prosody Control of End-to-end Speech Synthesis. [ICASSP 2019] [CoRR 2018]

• Xin Wang, Shinji Takaki, Junichi Yamagishi. Neural Source-filter-based Waveform Model for Statistical Parametric Speech Synthesis. [ICASSP 2019] [CoRR 2018]

  • TTS Spectrogram-to-waveform
  • Use source module to generate sine-based excitation signal, then use filter module to transform excitation signal into waveform.
  • Faster than autoregressive models (e.g., WaveNet).
  • Simpler than non-autoregressive models (e.g., Parallel WaveNet), since it does not need to employ complicated training method (e.g., distilling).

Poster 20: Speech Synthesis II

• Yusuke Yasuda, Xin Wang, Shinji Takaki, Junichi Yamagishi. Investigation of Enhanced Tacotron Text-to-speech Synthesis Systems with Self-attention for Pitch Accent Language. [ICASSP 2019] [CoRR 2018]

• Shan Yang, Heng Lu, Shiying Kang, Lei Xie, Dong Yu. Enhancing Hybrid Self-attention Structure with Relative-position-aware Bias for Speech Synthesis. [ICASSP 2019]

• Lauri Juvela, Bajibabu Bollepalli, Junichi Yamagishi, Paavo Alku. Waveform Generation for Text-to-speech Synthesis Using Pitch-synchronous Multi-scale Generative Adversarial Networks. [ICASSP 2019] [CoRR 2018]

• Kohki Mametani, Tsuneo Kato, Seiichi Yamamoto. Investigating Context Features Hidden in End-to-end TTS. [ICASSP 2019] [CoRR 2018]

• Éva Székely, Gustav Eje Henter, Joakim Gustafson. Casting to Corpus: Segmenting and Selecting Spontaneous Dialogue for TTS with a CNN-LSTM Speaker-dependent Breath Detector. [ICASSP 2019]

• Ruibo Fu, Jianhua Tao, Zhengqi Wen, Yibin Zheng. Phoneme Dependent Speaker Embedding and Model Factorization for Multi-speaker Speech Synthesis and Adaptation. [ICASSP 2019]

• Yuewen Cao, Xixin Wu, Songxiang Liu, Jianwei Yu, Xu Li, Zhiyong Wu, Xunying Liu, Helen Meng. End-to-end Code-switched TTS with Mix of Monolingual Recordings. [ICASSP 2019]

• Yu-An Chung, Yuxuan Wang, Wei-Ning Hsu, Yu Zhang, RJ Skerry-Ryan. Semi-supervised Training for Improving Data Efficiency in End-to-end Speech Synthesis. [ICASSP 2019] [CoRR 2018]

• Ya-Jie Zhang, Shifeng Pan, Lei He, Zhen-Hua Ling. Learning Latent Representations for Style Control and Transfer in End-to-end Speech Synthesis. [ICASSP 2019] [CoRR 2018]

• Heejin Choi, Sangjun Park, Jinuk Park, Minsoo Hahn. Multi-speaker Emotional Acoustic Modeling for CNN-based Speech Synthesis. [ICASSP 2019]

• Yukiya Hono, Kei Hashimoto, Keiichiro Oura, Yoshihiko Nankaku, Keiichi Tokuda. Singing Voice Synthesis Based on Generative Adversarial Networks. [ICASSP 2019]

• Kantapon Kaewtip, Fernando Villavicencio, Fang-Yu Kuo, Mark Harvilla, Iris Ouyang, Pierre Lanchantin. Enhanced Virtual Singers Generation by Incorporating Singing Dynamics to Personalized Text-to-speech-to-singing. [ICASSP 2019]

• Takuma Okamoto, Tomoki Toda, Yoshinori Shiga, Hisashi Kawai. Investigations of Real-time Gaussian Fftnet and Parallel Wavenet Neural Vocoders with Simple Acoustic Features. [ICASSP 2019]

Poster 22: Speech Synthesis III

• Yang Ai, Jing-Xuan Zhang, Liang Chen, Zhen-Hua Ling. DNN-based Spectral Enhancement for Neural Waveform Generators with Low-bit Quantization. [ICASSP 2019]

• Miraç Göksu Öztürk, Okan Ulusoy, Cenk Demiroglu. DNN-based Speaker-adaptive Postfiltering with Limited Adaptation Data for Statistical Speech Synthesis Systems. [ICASSP 2019]

• Chunhui Lu, Pengyuan Zhang, Yonghong Yan. Self-attention Based Prosodic Boundary Prediction for Chinese Speech Synthesis. [ICASSP 2019]

• François Marelli, Bastian Schnell, Hervé Bourlard, Thierry Dutoit, Philip N. Garner. An End-to-end Network to Synthesize Intonation Using a Generalized Command Response Model. [ICASSP 2019]

• Oliver Watts, Cassia Valentini-Botinhao, Simon King. Speech Waveform Reconstruction Using Convolutional Neural Networks with Noise and Periodic Inputs. [ICASSP 2019]

• Yanfeng Lu, Minghui Dong, Ying Chen. Implementing Prosodic Phrasing in Chinese End-to-end Speech Synthesis. [ICASSP 2019]

• Mu Wang, Xixin Wu, Zhiyong Wu, Shiyin Kang, Deyi Tuo, Guangzhi Li, Dan Su, Dong Yu, Helen Meng. Quasi-fully Convolutional Neural Network with Variational Inference for Speech Synthesis. [ICASSP 2019]

• Shinji Takaki, Toru Nakashika, Xin Wang, Junichi Yamagishi. STFT Spectral Loss for Training a Neural Speech Waveform Model. [ICASSP 2019] [CoRR 2018]

• Hiroki Tamaru, Yuki Saito, Shinnosuke Takamichi, Tomoki Koriyama, Hiroshi Saruwatari. Generative Moment Matching Network-based Random Modulation Post-filter for DNN-based Singing Voice Synthesis and Neural Double-tracking. [ICASSP 2019] [CoRR 2019]

• Javier Latorre, Jakub Lachowicz, Jaime Lorenzo-Trueba, Thomas Merritt, Thomas Drugman, Srikanth Ronanki, Viacheslav Klimkov. Effect of Data Reduction on Sequence-to-sequence Neural TTS. [ICASSP 2019] [CoRR 2018]