今日论文合集：cs.SD语音16篇，eess.AS音频处理16篇。
cs.SD语音

【1】 Deepfake audio as a data augmentation technique for training automatic speech to text transcription models标题：作为训练自动语音到文本转录模型的数据增强技术的深度伪音频作者：Alexandre R. Ferreira,Cláudio E. C. Campelo机构：Systems and Computing Department, Federal University of Campina Grande (UFCG), Campina Grande, Brazil, Cl´audio E. C. Campelo备注：9 pages, 6 figures, 7 tables链接：http://arxiv.org/pdf/2309.12802v1

【2】 CrossSinger: A Cross-Lingual Multi-Singer High-Fidelity Singing Voice Synthesizer Trained on Monolingual Singers标题：CrossSinger：一个针对单语歌手的跨语言多歌手高保真歌声合成器作者：Xintong Wang,Chang Zeng,Jun Chen,Chunhui Wang机构：Beijing Bombax XiaoIce Technology Co., Ltd, China, National Institute of Informatics, Japan ,SOKENDAI, Japan, Shenzhen International Graduate School, Tsinghua University, Shenzhen, China备注：Accepted by ASRU2023链接：http://arxiv.org/pdf/2309.12672v1
【3】 Profile-Error-Tolerant Target-Speaker Voice Activity Detection标题：轮廓容错目标说话人语音活动检测作者：Dongmei Wang,Xiong Xiao,Naoyuki Kanda,Midia Yousefi,Takuya Yoshioka,Jian Wu机构：Microsoft, One Microsoft Way, Redmond, WA, USA备注：Submission for ICASSP 2024链接：http://arxiv.org/pdf/2309.12521v1【4】 Memory-augmented conformer for improved end-to-end long-form ASR标题：用于改进的端到端长形ASR的存储器扩展整形器作者：Carlos Carvalho,Alberto Abad机构：INESC-ID, Lisbon, Portugal, Instituto Superior T´ecnico, Universidade de Lisboa, PortugalJournal-ref：Proc. INTERSPEECH 2023, 2218--2222链接：http://arxiv.org/pdf/2309.13029v1【5】 Dynamic ASR Pathways: An Adaptive Masking Approach Towards Efficient Pruning of A Multilingual ASR Model标题：动态ASR路径：一种多语言ASR模型高效剪枝的自适应掩蔽方法作者：Jiamin Xie,Ke Li,Jinxi Guo,Andros Tjandra,Yuan Shangguan,Leda Sari,Chunyang Wu,Junteng Jia,Jay Mahadeokar,Ozlem Kalinli机构：Center for Robust Speech Systems (CRSS), University of Texas at Dallas, USA, Meta AI, USA链接：http://arxiv.org/pdf/2309.13018v1【6】 Massive End-to-end Models for Short Search Queries标题：针对短搜索查询的大规模端到端模型作者：Weiran Wang,Rohit Prabhavalkar,Dongseong Hwang,Qiujia Li,Khe Chai Sim,Bo Li,James Qin,Xingyu Cai,Adam Stooke,Zhong Meng,CJ Zheng,Yanzhang He,Tara Sainath,Pedro Moreno Mengibar机构：Google LLC链接：http://arxiv.org/pdf/2309.12963v1【7】 VIC-KD: Variance-Invariance-Covariance Knowledge Distillation to Make Keyword Spotting More Robust Against Adversarial Attacks标题：VIC-KD：VARIANCE-INVARIANCE-COVARIANCE知识精馏，使关键词识别对敌方攻击更具鲁棒性作者：Heitor R. Guimarães,Arthur Pimentel,Anderson Avila,Tiago H. Falk机构：Institut national de la recherche scientifique (INRS-EMT), Universit´e du Qu´ebec, Montr´eal, Canada, and, INRS-UQO Mixed Research Unit on Cybersecurity, Gatineau,Qu´ebec, Canada备注：Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12914v1【8】 DurIAN-E: Duration Informed Attention Network For Expressive Text-to-Speech Synthesis标题：榴莲-E：面向表现性文语合成的时长感知注意网络作者：Yu Gu,Yianrao Bian,Guangzhi Lei,Chao Weng,Dan Su机构：Tencent AI Lab链接：http://arxiv.org/pdf/2309.12792v1【9】 A Study on Incorporating Whisper for Robust Speech Assessment标题：结合Whisper进行稳健语音评估的研究作者：Ryandhimas E. Zezario,Yu-Wen Chen,Yu Tsao,Szu-Wei Fu,Hsin-Min Wang,Chiou-Shann Fuh机构：National Taiwan University ,Academia Sinica ,Columbia University ,NVIDIA链接：http://arxiv.org/pdf/2309.12766v1【10】 Reduce, Reuse, Recycle: Is Perturbed Data better than Other Language augmentation for Low Resource Self-Supervised Speech Models标题：缩减、重用、再循环：对于低资源自监督语音模型，扰动数据是否比其他语言增强更好作者：Asad Ullah,Alessandro Ragano,Andrew Hines机构：School of Computer Science, University College Dublin, Ireland备注：5 pages, 4 figures, ICASSP24链接：http://arxiv.org/pdf/2309.12763v1【11】 Unsupervised Representations Improve Supervised Learning in Speech Emotion Recognition标题：无监督表征改进语音情感识别中的监督学习作者：Amirali Soltani Tehrani,Niloufar Faridani,Ramin Toosi机构：Electrical Engineering Faculty, University of Tehran, Tehran, Iran链接：http://arxiv.org/pdf/2309.12714v1【12】 Big model only for hard audios: Sample dependent Whisper model selection for efficient inferences标题：仅适用于硬音频的大模型：用于有效推理的依赖于样本的耳语模型选择作者：Hugo Malard,Salah Zaiem,Robin Algayres机构： ENS-PSL, Paris ,LTCI, T´el´ecom Paris, Institut Polytechnique de Paris, Inria, Paris备注：Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12712v1【13】 NTT speaker diarization system for CHiME-7: multi-domain, multi-microphone End-to-end and vector clustering diarization标题：适用于CHME-7的NTT说话人二值化系统：多域、多麦克风端到端和向量聚类二值化作者：Naohiro Tawara,Marc Delcroix,Atsushi Ando,Atsunori Ogawa机构：NTT Corporation, Japan备注：5 pages, 5 figures, Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12656v1【14】 SPGM: Prioritizing Local Features for enhanced speech separation performance标题：SPGM：优先处理局部特征以增强语音分离性能作者：Jia Qi Yip,Shengkui Zhao,Yukun Ma,Chongjia Ni,Chong Zhang,Hao Wang,Trung Hieu Nguyen,Kun Zhou,Dianwen Ng,Eng Siong Chng,Bin Ma链接：http://arxiv.org/pdf/2309.12608v1【15】 Sampling-Frequency-Independent Universal Sound Separation标题：与采样频率无关的通用声音分离作者：Tomohiko Nakamura,Kohei Yatabe机构：†National Institute of Advanced Industrial Science and Technology (AIST), Tokyo ,-, Japan, ‡Tokyo University of Agriculture and Technology (TUAT), Tokyo ,-, Japan备注：Submitted to ICASSP2024链接：http://arxiv.org/pdf/2309.12581v1【16】 ICASSP 2023 Acoustic Echo Cancellation Challenge标题：ICASSP 2023声学回波消除挑战赛作者：Ross Cutler,Ando Saabas,Tanel Parnamaa,Marju Purin,Evgenii Indenbom,Nicolae-Catalin Ristea,Jegor Gužvin,Hannes Gamper,Sebastian Braun,Robert Aichner机构：Aichner, Microsoft Corporation, Redmond, USA备注：arXiv admin note: substantial text overlap with arXiv:2202.13290, arXiv:2009.04972链接：http://arxiv.org/pdf/2309.12553v1
eess.AS音频处理

【1】 Memory-augmented conformer for improved end-to-end long-form ASR标题：用于改进的端到端长形ASR的存储器扩展整形器作者：Carlos Carvalho,Alberto Abad机构：INESC-ID, Lisbon, Portugal, Instituto Superior T´ecnico, Universidade de Lisboa, PortugalJournal-ref：Proc. INTERSPEECH 2023, 2218--2222链接：http://arxiv.org/pdf/2309.13029v1

【2】 Dynamic ASR Pathways: An Adaptive Masking Approach Towards Efficient Pruning of A Multilingual ASR Model标题：动态ASR路径：一种多语言ASR模型高效剪枝的自适应掩蔽方法作者：Jiamin Xie,Ke Li,Jinxi Guo,Andros Tjandra,Yuan Shangguan,Leda Sari,Chunyang Wu,Junteng Jia,Jay Mahadeokar,Ozlem Kalinli机构：Center for Robust Speech Systems (CRSS), University of Texas at Dallas, USA, Meta AI, USA链接：http://arxiv.org/pdf/2309.13018v1
【3】 Massive End-to-end Models for Short Search Queries标题：针对短搜索查询的大规模端到端模型作者：Weiran Wang,Rohit Prabhavalkar,Dongseong Hwang,Qiujia Li,Khe Chai Sim,Bo Li,James Qin,Xingyu Cai,Adam Stooke,Zhong Meng,CJ Zheng,Yanzhang He,Tara Sainath,Pedro Moreno Mengibar机构：Google LLC链接：http://arxiv.org/pdf/2309.12963v1【4】 VIC-KD: Variance-Invariance-Covariance Knowledge Distillation to Make Keyword Spotting More Robust Against Adversarial Attacks标题：VIC-KD：VARIANCE-INVARIANCE-COVARIANCE知识精馏，使关键词识别对敌方攻击更具鲁棒性作者：Heitor R. Guimarães,Arthur Pimentel,Anderson Avila,Tiago H. Falk机构：Institut national de la recherche scientifique (INRS-EMT), Universit´e du Qu´ebec, Montr´eal, Canada, and, INRS-UQO Mixed Research Unit on Cybersecurity, Gatineau,Qu´ebec, Canada备注：Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12914v1【5】 DurIAN-E: Duration Informed Attention Network For Expressive Text-to-Speech Synthesis标题：榴莲-E：面向表现性文语合成的时长感知注意网络作者：Yu Gu,Yianrao Bian,Guangzhi Lei,Chao Weng,Dan Su机构：Tencent AI Lab链接：http://arxiv.org/pdf/2309.12792v1
【6】 A Study on Incorporating Whisper for Robust Speech Assessment标题：结合Whisper进行稳健语音评估的研究作者：Ryandhimas E. Zezario,Yu-Wen Chen,Yu Tsao,Szu-Wei Fu,Hsin-Min Wang,Chiou-Shann Fuh机构：National Taiwan University ,Academia Sinica ,Columbia University ,NVIDIA链接：http://arxiv.org/pdf/2309.12766v1【7】 Reduce, Reuse, Recycle: Is Perturbed Data better than Other Language augmentation for Low Resource Self-Supervised Speech Models标题：缩减、重用、再循环：对于低资源自监督语音模型，扰动数据是否比其他语言增强更好作者：Asad Ullah,Alessandro Ragano,Andrew Hines机构：School of Computer Science, University College Dublin, Ireland备注：5 pages, 4 figures, ICASSP24链接：http://arxiv.org/pdf/2309.12763v1【8】 Unsupervised Representations Improve Supervised Learning in Speech Emotion Recognition标题：无监督表征改进语音情感识别中的监督学习作者：Amirali Soltani Tehrani,Niloufar Faridani,Ramin Toosi机构：Electrical Engineering Faculty, University of Tehran, Tehran, Iran链接：http://arxiv.org/pdf/2309.12714v1【9】 Big model only for hard audios: Sample dependent Whisper model selection for efficient inferences标题：仅适用于硬音频的大模型：用于有效推理的依赖于样本的耳语模型选择作者：Hugo Malard,Salah Zaiem,Robin Algayres机构： ENS-PSL, Paris ,LTCI, T´el´ecom Paris, Institut Polytechnique de Paris, Inria, Paris备注：Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12712v1【10】 NTT speaker diarization system for CHiME-7: multi-domain, multi-microphone End-to-end and vector clustering diarization标题：适用于CHME-7的NTT说话人二值化系统：多域、多麦克风端到端和向量聚类二值化作者：Naohiro Tawara,Marc Delcroix,Atsushi Ando,Atsunori Ogawa机构：NTT Corporation, Japan备注：5 pages, 5 figures, Submitted to ICASSP 2024链接：http://arxiv.org/pdf/2309.12656v1【11】 SPGM: Prioritizing Local Features for enhanced speech separation performance标题：SPGM：优先处理局部特征以增强语音分离性能作者：Jia Qi Yip,Shengkui Zhao,Yukun Ma,Chongjia Ni,Chong Zhang,Hao Wang,Trung Hieu Nguyen,Kun Zhou,Dianwen Ng,Eng Siong Chng,Bin Ma链接：http://arxiv.org/pdf/2309.12608v1

【12】 Sampling-Frequency-Independent Universal Sound Separation标题：与采样频率无关的通用声音分离作者：Tomohiko Nakamura,Kohei Yatabe机构：†National Institute of Advanced Industrial Science and Technology (AIST), Tokyo ,-, Japan, ‡Tokyo University of Agriculture and Technology (TUAT), Tokyo ,-, Japan备注：Submitted to ICASSP2024链接：http://arxiv.org/pdf/2309.12581v1【13】 ICASSP 2023 Acoustic Echo Cancellation Challenge标题：ICASSP 2023声学回波消除挑战赛作者：Ross Cutler,Ando Saabas,Tanel Parnamaa,Marju Purin,Evgenii Indenbom,Nicolae-Catalin Ristea,Jegor Gužvin,Hannes Gamper,Sebastian Braun,Robert Aichner机构：Aichner, Microsoft Corporation, Redmond, USA备注：arXiv admin note: substantial text overlap with arXiv:2202.13290, arXiv:2009.04972链接：http://arxiv.org/pdf/2309.12553v1【14】 Deepfake audio as a data augmentation technique for training automatic speech to text transcription models标题：作为训练自动语音到文本转录模型的数据增强技术的深度伪音频作者：Alexandre R. Ferreira,Cláudio E. C. Campelo机构：Systems and Computing Department, Federal University of Campina Grande (UFCG), Campina Grande, Brazil, Cl´audio E. C. Campelo备注：9 pages, 6 figures, 7 tables链接：http://arxiv.org/pdf/2309.12802v1【15】 CrossSinger: A Cross-Lingual Multi-Singer High-Fidelity Singing Voice Synthesizer Trained on Monolingual Singers标题：CrossSinger：一个针对单语歌手的跨语言多歌手高保真歌声合成器作者：Xintong Wang,Chang Zeng,Jun Chen,Chunhui Wang机构：Beijing Bombax XiaoIce Technology Co., Ltd, China, National Institute of Informatics, Japan ,SOKENDAI, Japan, Shenzhen International Graduate School, Tsinghua University, Shenzhen, China备注：Accepted by ASRU2023链接：http://arxiv.org/pdf/2309.12672v1【16】 Profile-Error-Tolerant Target-Speaker Voice Activity Detection标题：轮廓容错目标说话人语音活动检测作者：Dongmei Wang,Xiong Xiao,Naoyuki Kanda,Midia Yousefi,Takuya Yoshioka,Jian Wu机构：Microsoft, One Microsoft Way, Redmond, WA, USA备注：Submission for ICASSP 2024链接：http://arxiv.org/pdf/2309.12521v1