今日论文合集：cs.SD语音7篇，eess.AS音频处理10篇。
cs.SD语音

【1】A Scalable Framework for Automatic Playlist Continuation on Music Streaming Services标题：一种可扩展的音乐流媒体服务播放列表自动延续框架链接：http://arxiv.org/pdf/2304.09061v1

作者：Walid Bendada,Guillaume Salha-Galvan,Thomas Bouabça,Tristan Cazenave机构：Deezer Research, LAMSADE, Université Paris Dauphine, PSL备注：Accepted as a Full Paper at the SIGIR 2023 conference摘要：音乐流媒体服务通常旨在为用户推荐歌曲以扩展他们在这些服务上创建的播放列表。然而，扩展播放列表同时保留其音乐特性并匹配用户偏好仍然是一项具有挑战性的任务，通常被称为自动播放列表延续（APC）。此外，虽然这些服务通常需要在具有数百万候选者的大型目录中实时选择最佳歌曲进行推荐，但最近对APC的研究主要集中在可扩展性保证很少的模型上，并在相对较小的数据集上进行评估。在本文中，我们介绍了一个通用的框架，以建立可扩展的，但有效的APC模型的大规模应用程序。基于先表示再聚合的策略，它通过设计确保可扩展性，同时保持足够的灵活性，以结合广泛的表示学习和序列建模技术，例如，基于Transformers。我们通过对Spotify的百万播放列表数据集（MPD）（APC最大的公共数据集）进行深入的实验验证，证明了该框架的相关性。我们还描述了在2022年，我们如何成功地利用这个框架来改进Deezer生产中的APC。我们报告的结果，从一个大规模的在线A B测试，这项服务，强调我们的方法在这样一个现实世界中的应用程序的实际影响。摘要：Music streaming services often aim to recommend songs for users to extend the playlists they have created on these services. However, extending playlists while preserving their musical characteristics and matching user preferences remains a challenging task, commonly referred to as Automatic Playlist Continuation (APC). Besides, while these services often need to select the best songs to recommend in real-time and among large catalogs with millions of candidates, recent research on APC mainly focused on models with few scalability guarantees and evaluated on relatively small datasets. In this paper, we introduce a general framework to build scalable yet effective APC models for large-scale applications. Based on a represent-then-aggregate strategy, it ensures scalability by design while remaining flexible enough to incorporate a wide range of representation learning and sequence modeling techniques, e.g., based on Transformers. We demonstrate the relevance of this framework through in-depth experimental validation on Spotify's Million Playlist Dataset (MPD), the largest public dataset for APC. We also describe how, in 2022, we successfully leveraged this framework to improve APC in production on Deezer. We report results from a large-scale online A B test on this service, emphasizing the practical impact of our approach in such a real-world application.

【2】From Words to Music: A Study of Subword Tokenization Techniques in Symbolic Music Generation标题：从文字到音乐：符号音乐生成中的子词标记化技术研究链接：http://arxiv.org/pdf/2304.08953v1

作者：Adarsh Kumar,Pedro Sarmento机构：Indian Institute of Technology Kharagpur, India, Queen Mary University of London, UK摘要：子词标记化在基于Transformer模型的基于文本的自然语言处理（NLP）任务中取得了广泛的成功。随着Transformer模型在符号音乐相关研究中越来越受欢迎，研究符号音乐域中的子词标记化的功效势在必行。在本文中，我们探讨了子词标记化技术，如字节对编码（BPE），在符号音乐生成及其对生成的歌曲的整体结构的影响。我们的实验基于三种类型的MIDI数据集：仅单音轨旋律、具有单个乐器的多音轨、以及多音轨和多乐器。我们应用子词标记后的音乐标记化方案，发现它能够生成较长的歌曲在同一时间，并提高了整体结构的客观指标，如结构指标（SI），音高类熵等方面所生成的音乐，我们还比较了两个子词标记化方法，BPE和Unigram，并观察到这两种方法导致一致的改进。我们的研究表明，子词标记化是一种很有前途的符号音乐生成技术，并可能对音乐创作产生更广泛的影响，特别是在涉及复杂数据的情况下，如多轨歌曲。摘要：Subword tokenization has been widely successful in text-based natural language processing (NLP) tasks with Transformer-based models. As Transformer models become increasingly popular in symbolic music-related studies, it is imperative to investigate the efficacy of subword tokenization in the symbolic music domain. In this paper, we explore subword tokenization techniques, such as byte-pair encoding (BPE), in symbolic music generation and its impact on the overall structure of generated songs. Our experiments are based on three types of MIDI datasets: single track-melody only, multi-track with a single instrument, and multi-track and multi-instrument. We apply subword tokenization on post-musical tokenization schemes and find that it enables the generation of longer songs at the same time and improves the overall structure of the generated music in terms of objective metrics like structure indicator (SI), Pitch Class Entropy, etc. We also compare two subword tokenization methods, BPE and Unigram, and observe that both methods lead to consistent improvements. Our study suggests that subword tokenization is a promising technique for symbolic music generation and may have broader implications for music composition, particularly in cases involving complex data such as multi-track songs.

【3】Towards the Transferable Audio Adversarial Attack via Ensemble Methods标题：基于集成方法的可转移音频对抗攻击链接：http://arxiv.org/pdf/2304.08811v1

作者：Feng Guo,Zheng Sun,Yuxuan Chen,Lei Ju机构：cn 1School of Cyber Science and Technology, Shandong University备注：Submitted to Cybersecurity journal 2023摘要：近年来，深度学习（DL）模型在许多领域取得了重大进展，例如自动驾驶，面部识别和语音识别。然而，深度学习模型对对抗性攻击的脆弱性引起了社区的严重关注，因为它们的鲁棒性和泛化能力不足。此外，可转移攻击已经成为黑盒攻击的一种重要方法。在这项工作中，我们探讨了影响基于DL的语音识别中对抗性示例（AE）可转移性的潜在因素。我们还讨论了不同的DL系统的脆弱性和决策边界的不规则性。我们的研究结果表明，语音和图像之间的AE的可转移性显着的差异，与数据的相关性低的图像，但在语音识别相反。出于辍学为基础的集成方法，我们提出了随机梯度合奏和动态梯度加权合奏，我们评估合奏的影响AE的可转移性。结果表明，由这两种方法创建的AE对于转移到黑盒API是有效的。摘要：In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufficient robustness and generalization. Also, transferable attacks have become a prominent method for black-box attacks. In this work, we explore the potential factors that impact adversarial examples (AEs) transferability in DL-based speech recognition. We also discuss the vulnerability of different DL systems and the irregular nature of decision boundaries. Our results show a remarkable difference in the transferability of AEs between speech and images, with the data relevance being low in images but opposite in speech recognition. Motivated by dropout-based ensemble approaches, we propose random gradient ensembles and dynamic gradient-weighted ensembles, and we evaluate the impact of ensembles on the transferability of AEs. The results show that the AEs created by both approaches are valid for transfer to the black box API.

【4】A Voice Disease Detection Method Based on MFCCs and Shallow CNN标题：一种基于MFCCS和浅层CNN的语音疾病检测方法链接：http://arxiv.org/pdf/2304.08708v1

作者：Xiaoping Xie,Hao Cai,Can Li,Fei Ding机构：We cooperated with Xiangya Hospital of Central South University to collect voice samples from sixty-one different patients, Author was with Rice University, He is now with the Department of Physics, Colorado State University摘要：嗓音疾病的发病率呈逐年上升趋势。利用软件进行远程诊断是一种技术发展趋势，具有重要的实用价值。在嗓音疾病中，引起声音嘶哑的常见疾病包括痉挛性发声障碍、声带麻痹、声带小结和声带息肉。本文提出了一种可广泛应用于临床的嗓音疾病检测方法。我们与中南大学湘雅医院合作，收集了61名不同患者的声音样本。提取Mel频率倒谱系数（MFCC）参数作为输入特征，以数据的形式描述语音。提出了一种结合MFCC参数和单卷积层CNN的创新模型，用于快速计算和分类。我们达到的最高准确率为92%，这是完全领先于原来的研究成果和国际先进水平。并利用高级语音功能评估数据库（AVFAD）对该方法的泛化能力进行了测试，准确率达到98%。在临床和标准数据集上的实验表明，对于嗓音疾病的病理检测，该方法在准确率和计算效率上都有很大的提高。摘要：The incidence rate of voice diseases is increasing year by year. The use of software for remote diagnosis is a technical development trend and has important practical value. Among voice diseases, common diseases that cause hoarseness include spasmodic dysphonia, vocal cord paralysis, vocal nodule, and vocal cord polyp. This paper presents a voice disease detection method that can be applied in a wide range of clinical. We cooperated with Xiangya Hospital of Central South University to collect voice samples from sixty-one different patients. The Mel Frequency Cepstrum Coefficient (MFCC) parameters are extracted as input features to describe the voice in the form of data. An innovative model combining MFCC parameters and single convolution layer CNN is proposed for fast calculation and classification. The highest accuracy we achieved was 92%, it is fully ahead of the original research results and internationally advanced. And we use Advanced Voice Function Assessment Databases (AVFAD) to evaluate the generalization ability of the method we proposed, which achieved an accuracy rate of 98%. Experiments on clinical and standard datasets show that for the pathological detection of voice diseases, our method has greatly improved in accuracy and computational efficiency.

【5】NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers标题：NaturalSpeech 2：潜在扩散模型是自然和零射语音和歌唱合成器链接：http://arxiv.org/pdf/2304.09116v1

作者：Kai Shen,Zeqian Ju,Xu Tan,Yanqing Liu,Yichong Leng,Lei He,Tao Qin,Sheng Zhao,Jiang Bian机构：Microsoft Research Asia & Microsoft Azure Speech备注：A large-scale text-to-speech and singing voice synthesis system with latent diffusion models摘要：将文本到语音（TTS）缩放到大规模、多说话者和野外数据集对于捕获人类语音的多样性（诸如说话者身份、韵律和风格（例如，语音识别））是重要的。唱歌）。目前的大型TTS系统通常将语音量化为离散的令牌，并使用语言模型来逐个生成这些令牌，这遭受不稳定的韵律，单词跳过 重复问题，以及较差的语音质量。在本文中，我们开发的NaturalSpeech 2，TTS系统，利用神经音频编解码器与残余向量量化器得到量化的潜向量，并使用扩散模型来生成这些潜向量的文本输入的条件。为了提高zero-shot能力，这是重要的，以实现多样化的语音合成，我们设计了一个语音提示机制，以促进在上下文学习的扩散模型和持续时间 音高预测。我们将NaturalSpeech 2扩展到具有44 K小时语音和唱歌数据的大规模数据集，并在看不见的扬声器上评估其语音质量。NaturalSpeech 2在韵律 音色相似性，鲁棒性和zero-shot设置中的语音质量方面大大优于以前的TTS系统，并且仅使用语音提示进行新颖的零拍歌唱合成。音频样本可在www.example.com上获得https: speechresearch.github.io naturalspeech2。摘要：Scaling text-to-speech (TTS) to large-scale, multi-speaker, and in-the-wild datasets is important to capture the diversity in human speech such as speaker identities, prosodies, and styles (e.g., singing). Current large TTS systems usually quantize speech into discrete tokens and use language models to generate these tokens one by one, which suffer from unstable prosody, word skipping repeating issue, and poor voice quality. In this paper, we develop NaturalSpeech 2, a TTS system that leverages a neural audio codec with residual vector quantizers to get the quantized latent vectors and uses a diffusion model to generate these latent vectors conditioned on text input. To enhance the zero-shot capability that is important to achieve diverse speech synthesis, we design a speech prompting mechanism to facilitate in-context learning in the diffusion model and the duration pitch predictor. We scale NaturalSpeech 2 to large-scale datasets with 44K hours of speech and singing data and evaluate its voice quality on unseen speakers. NaturalSpeech 2 outperforms previous TTS systems by a large margin in terms of prosody timbre similarity, robustness, and voice quality in a zero-shot setting, and performs novel zero-shot singing synthesis with only a speech prompt. Audio samples are available at https: speechresearch.github.io naturalspeech2.

【6】Neural Speech Enhancement with Very Low Algorithmic Latency and Complexity via Integrated Full- and Sub-Band Modeling标题：集成全带和子带建模的神经语音增强算法延时和复杂度极低链接：http://arxiv.org/pdf/2304.08707v1

作者：Zhong-Qiu Wang,Samuele Cornell,Shukjae Choi,Younglo Lee,Byeong-Yeol Kim,Shinji Watanabe机构：Language Technologies Institute, Carnegie Mellon University, Pittsburgh, USA, Italy, Hyundai Motor Group and ,dot Inc., Seoul, Korea备注：in ICASSP 2023摘要：我们提出了FSB-LSTM，一种新的基于长短期记忆（LSTM）的架构，集成了全频带和子频带（FSB）建模，用于短时傅里叶变换（STFT）域中的单通道和多通道语音增强。该模型维护了一条信息高速公路，以使过完备输入表示流经多个FSB-LSTM模块。每个FSB-LSTM模块由全带块和子带块组成，全带块用于对所有频率处的频谱-时间模式进行建模，子带块用于对每个子带内的模式进行建模，其中两个块中的每一个将下采样表示作为输入，并返回上采样的判别表示以经由残差连接添加到块输入。该模型被设计为具有低的算法复杂度，一个小的运行时缓冲区和一个非常低的算法延迟，在同一时间产生一个强大的增强性能的噪声混响语音增强任务，即使跳的大小是低至2 $ms。摘要：We propose FSB-LSTM, a novel long short-term memory (LSTM) based architecture that integrates full- and sub-band (FSB) modeling, for single- and multi-channel speech enhancement in the short-time Fourier transform (STFT) domain. The model maintains an information highway to flow an over-complete input representation through multiple FSB-LSTM modules. Each FSB-LSTM module consists of a full-band block to model spectro-temporal patterns at all frequencies and a sub-band block to model patterns within each sub-band, where each of the two blocks takes a down-sampled representation as input and returns an up-sampled discriminative representation to be added to the block input via a residual connection. The model is designed to have a low algorithmic complexity, a small run-time buffer and a very low algorithmic latency, at the same time producing a strong enhancement performance on a noisy-reverberant speech enhancement task even if the hop size is as low as $2$ ms.

【7】How Tiny Can Analog Filterbank Features Be Made for Ultra-low-power On-device Keyword Spotting?标题：模拟滤波器组功能可用于超低功耗设备上的关键字检测的微小程度？链接：http://arxiv.org/pdf/2304.08541v1

作者：Subhajit Ray,Xinghua Sun,Nolan Tremelling,Maria Gordiyenko,Peter Kinget机构：Columbia University, New York, United States, Barnard College备注：Accepted as a full paper by the TinyML Research Symposium 2023摘要：模拟特征提取是用于在设备上关键字定位系统中实现前端特征提取器的功率高效且重新出现的信号处理范例。尽管其功率效率和重新出现，但对于其关键块（模拟滤波器组）的架构参数应设置为什么值，几乎没有共识，即使它们强烈影响功耗。为了建立共识并接近基本功耗限制，我们通过模拟发现，通过仔细选择其架构参数，典型的最先进模拟滤波器组的功率可以降低33.6倍，而通过后端神经网络的下游10字关键字定位准确性仅牺牲1.8%。摘要：Analog feature extraction is a power-efficient and re-emerging signal processing paradigm for implementing the front-end feature extractor in on device keyword-spotting systems. Despite its power efficiency and re-emergence, there is little consensus on what values the architectural parameters of its critical block, the analog filterbank, should be set to, even though they strongly influence power consumption. Towards building consensus and approaching fundamental power consumption limits, we find via simulation that through careful selection of its architectural parameters, the power of a typical state-of-the-art analog filterbank could be reduced by 33.6x, while sacrificing only 1.8% in downstream 10-word keyword spotting accuracy through a back-end neural network.
eess.AS音频处理

【1】NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers标题：NaturalSpeech 2：潜在扩散模型是自然和零射语音和歌唱合成器链接：http://arxiv.org/pdf/2304.09116v1

*与cs.SD语音【5】为同一篇

作者：Kai Shen,Zeqian Ju,Xu Tan,Yanqing Liu,Yichong Leng,Lei He,Tao Qin,Sheng Zhao,Jiang Bian机构：Microsoft Research Asia & Microsoft Azure Speech备注：A large-scale text-to-speech and singing voice synthesis system with latent diffusion models摘要：将文本到语音（TTS）缩放到大规模、多说话者和野外数据集对于捕获人类语音的多样性（诸如说话者身份、韵律和风格（例如，语音识别））是重要的。唱歌）。目前的大型TTS系统通常将语音量化为离散的令牌，并使用语言模型来逐个生成这些令牌，这遭受不稳定的韵律，单词跳过 重复问题，以及较差的语音质量。在本文中，我们开发的NaturalSpeech 2，TTS系统，利用神经音频编解码器与残余向量量化器得到量化的潜向量，并使用扩散模型来生成这些潜向量的文本输入的条件。为了提高zero-shot能力，这是重要的，以实现多样化的语音合成，我们设计了一个语音提示机制，以促进在上下文学习的扩散模型和持续时间 音高预测。我们将NaturalSpeech 2扩展到具有44 K小时语音和唱歌数据的大规模数据集，并在看不见的扬声器上评估其语音质量。NaturalSpeech 2在韵律 音色相似性，鲁棒性和zero-shot设置中的语音质量方面大大优于以前的TTS系统，并且仅使用语音提示进行新颖的零拍歌唱合成。音频样本可在www.example.com上获得https: speechresearch.github.io naturalspeech2。摘要：Scaling text-to-speech (TTS) to large-scale, multi-speaker, and in-the-wild datasets is important to capture the diversity in human speech such as speaker identities, prosodies, and styles (e.g., singing). Current large TTS systems usually quantize speech into discrete tokens and use language models to generate these tokens one by one, which suffer from unstable prosody, word skipping repeating issue, and poor voice quality. In this paper, we develop NaturalSpeech 2, a TTS system that leverages a neural audio codec with residual vector quantizers to get the quantized latent vectors and uses a diffusion model to generate these latent vectors conditioned on text input. To enhance the zero-shot capability that is important to achieve diverse speech synthesis, we design a speech prompting mechanism to facilitate in-context learning in the diffusion model and the duration pitch predictor. We scale NaturalSpeech 2 to large-scale datasets with 44K hours of speech and singing data and evaluate its voice quality on unseen speakers. NaturalSpeech 2 outperforms previous TTS systems by a large margin in terms of prosody timbre similarity, robustness, and voice quality in a zero-shot setting, and performs novel zero-shot singing synthesis with only a speech prompt. Audio samples are available at https: speechresearch.github.io naturalspeech2.

【2】Array Configuration-Agnostic Personal Voice Activity Detection Based on Spatial Coherence标题：基于空间相干性的阵列配置无关的个人语音活动检测链接：http://arxiv.org/pdf/2304.08887v1

作者：Yicheng Hsu,Mingsian R. Bai机构：Department of Power Mechanical Engineering, National Tsing Hua University, Section , Kuang-Fu Road, Hsinchu, Taiwan , Department of Power Mechanical Engineering and Electrical Engineering, National Tsing Hua备注：Accepted by INTER-NOISE 2023. arXiv admin note: text overlap with arXiv:2211.08748摘要：由于个人移动设备和智能扬声器的日益普及，个人语音活动检测已经受到越来越多的关注。PVAD通常是这些应用中语音增强和识别的一个组成部分，在这些应用中，轻量级信号处理仅针对目标用户启用。然而，在现实世界的情况下，检测性能可能会下降，因为竞争的扬声器，背景噪声，和混响。为了解决这个问题，我们提出了使用等效矩形带宽ERB缩放的空间相干性作为输入特征来训练阵列配置不可知的PVAD网络。而网络模型只需要112 k个参数，它表现出优异的检测性能和鲁棒性，在不利的声学条件。值得注意的是，所提出的ARCA-PVAD系统可扩展到阵列配置。实验结果表明，所提出的ARCA-PVAD系统在接收机工作特性曲线下的面积和等误码率方面超过基线，实现了优异的性能。摘要：Personal voice activity detection has received increased attention due to the growing popularity of personal mobile devices and smart speakers. PVAD is often an integral element to speech enhancement and recognition for these applications in which lightweight signal processing is only enabled for the target user. However, in real-world scenarios, the detection performance may degrade because of competing speakers, background noise, and reverberation. To address this problem, we proposed to use equivalent rectangular bandwidth ERB-scaled spatial coherence as the input feature to train an array configuration-agnostic PVAD network. Whereas the network model requires only 112k parameters, it exhibits excellent detection performance and robustness in adverse acoustic conditions. Notably, the proposed ARCA-PVAD system is scalable to array configurations. Experimental results have demonstrated the superior performance achieved by the proposed ARCA-PVAD system over a baseline in terms of the area under receiver operating characteristic curve and equal error rate.

【3】Neural Speech Enhancement with Very Low Algorithmic Latency and Complexity via Integrated Full- and Sub-Band Modeling标题：集成全带和子带建模的神经语音增强算法延时和复杂度极低链接：http://arxiv.org/pdf/2304.08707v1

*与cs.SD语音【6】为同一篇

作者：Zhong-Qiu Wang,Samuele Cornell,Shukjae Choi,Younglo Lee,Byeong-Yeol Kim,Shinji Watanabe机构：Language Technologies Institute, Carnegie Mellon University, Pittsburgh, USA, Italy, Hyundai Motor Group and ,dot Inc., Seoul, Korea备注：in ICASSP 2023摘要：我们提出了FSB-LSTM，一种新的基于长短期记忆（LSTM）的架构，集成了全频带和子频带（FSB）建模，用于短时傅里叶变换（STFT）域中的单通道和多通道语音增强。该模型维护了一条信息高速公路，以使过完备输入表示流经多个FSB-LSTM模块。每个FSB-LSTM模块由全带块和子带块组成，全带块用于对所有频率处的频谱-时间模式进行建模，子带块用于对每个子带内的模式进行建模，其中两个块中的每一个将下采样表示作为输入，并返回上采样的判别表示以经由残差连接添加到块输入。该模型被设计为具有低的算法复杂度，一个小的运行时缓冲区和一个非常低的算法延迟，在同一时间产生一个强大的增强性能的噪声混响语音增强任务，即使跳的大小是低至2 $ms。摘要：We propose FSB-LSTM, a novel long short-term memory (LSTM) based architecture that integrates full- and sub-band (FSB) modeling, for single- and multi-channel speech enhancement in the short-time Fourier transform (STFT) domain. The model maintains an information highway to flow an over-complete input representation through multiple FSB-LSTM modules. Each FSB-LSTM module consists of a full-band block to model spectro-temporal patterns at all frequencies and a sub-band block to model patterns within each sub-band, where each of the two blocks takes a down-sampled representation as input and returns an up-sampled discriminative representation to be added to the block input via a residual connection. The model is designed to have a low algorithmic complexity, a small run-time buffer and a very low algorithmic latency, at the same time producing a strong enhancement performance on a noisy-reverberant speech enhancement task even if the hop size is as low as $2$ ms.

【4】How Tiny Can Analog Filterbank Features Be Made for Ultra-low-power On-device Keyword Spotting?标题：模拟滤波器组功能可用于超低功耗设备上的关键字检测的微小程度？链接：http://arxiv.org/pdf/2304.08541v1

*与cs.SD语音【7】为同一篇

作者：Subhajit Ray,Xinghua Sun,Nolan Tremelling,Maria Gordiyenko,Peter Kinget机构：Columbia University, New York, United States, Barnard College备注：Accepted as a full paper by the TinyML Research Symposium 2023摘要：模拟特征提取是用于在设备上关键字定位系统中实现前端特征提取器的功率高效且重新出现的信号处理范例。尽管其功率效率和重新出现，但对于其关键块（模拟滤波器组）的架构参数应设置为什么值，几乎没有共识，即使它们强烈影响功耗。为了建立共识并接近基本功耗限制，我们通过模拟发现，通过仔细选择其架构参数，典型的最先进模拟滤波器组的功率可以降低33.6倍，而通过后端神经网络的下游10字关键字定位准确性仅牺牲1.8%。摘要：Analog feature extraction is a power-efficient and re-emerging signal processing paradigm for implementing the front-end feature extractor in on device keyword-spotting systems. Despite its power efficiency and re-emergence, there is little consensus on what values the architectural parameters of its critical block, the analog filterbank, should be set to, even though they strongly influence power consumption. Towards building consensus and approaching fundamental power consumption limits, we find via simulation that through careful selection of its architectural parameters, the power of a typical state-of-the-art analog filterbank could be reduced by 33.6x, while sacrificing only 1.8% in downstream 10-word keyword spotting accuracy through a back-end neural network.

【5】A Scalable Framework for Automatic Playlist Continuation on Music Streaming Services标题：一种可扩展的音乐流媒体服务播放列表自动延续框架链接：http://arxiv.org/pdf/2304.09061v1

*与cs.SD语音【1】为同一篇

作者：Walid Bendada,Guillaume Salha-Galvan,Thomas Bouabça,Tristan Cazenave机构：Deezer Research, LAMSADE, Université Paris Dauphine, PSL备注：Accepted as a Full Paper at the SIGIR 2023 conference摘要：音乐流媒体服务通常旨在为用户推荐歌曲以扩展他们在这些服务上创建的播放列表。然而，扩展播放列表同时保留其音乐特性并匹配用户偏好仍然是一项具有挑战性的任务，通常被称为自动播放列表延续（APC）。此外，虽然这些服务通常需要在具有数百万候选者的大型目录中实时选择最佳歌曲进行推荐，但最近对APC的研究主要集中在可扩展性保证很少的模型上，并在相对较小的数据集上进行评估。在本文中，我们介绍了一个通用的框架，以建立可扩展的，但有效的APC模型的大规模应用程序。基于先表示再聚合的策略，它通过设计确保可扩展性，同时保持足够的灵活性，以结合广泛的表示学习和序列建模技术，例如，基于Transformers。我们通过对Spotify的百万播放列表数据集（MPD）（APC最大的公共数据集）进行深入的实验验证，证明了该框架的相关性。我们还描述了在2022年，我们如何成功地利用这个框架来改进Deezer生产中的APC。我们报告的结果，从一个大规模的在线A B测试，这项服务，强调我们的方法在这样一个现实世界中的应用程序的实际影响。摘要：Music streaming services often aim to recommend songs for users to extend the playlists they have created on these services. However, extending playlists while preserving their musical characteristics and matching user preferences remains a challenging task, commonly referred to as Automatic Playlist Continuation (APC). Besides, while these services often need to select the best songs to recommend in real-time and among large catalogs with millions of candidates, recent research on APC mainly focused on models with few scalability guarantees and evaluated on relatively small datasets. In this paper, we introduce a general framework to build scalable yet effective APC models for large-scale applications. Based on a represent-then-aggregate strategy, it ensures scalability by design while remaining flexible enough to incorporate a wide range of representation learning and sequence modeling techniques, e.g., based on Transformers. We demonstrate the relevance of this framework through in-depth experimental validation on Spotify's Million Playlist Dataset (MPD), the largest public dataset for APC. We also describe how, in 2022, we successfully leveraged this framework to improve APC in production on Deezer. We report results from a large-scale online A B test on this service, emphasizing the practical impact of our approach in such a real-world application.

【6】From Words to Music: A Study of Subword Tokenization Techniques in Symbolic Music Generation标题：从文字到音乐：符号音乐生成中的子词标记化技术研究链接：http://arxiv.org/pdf/2304.08953v1

*与cs.SD语音【2】为同一篇

作者：Adarsh Kumar,Pedro Sarmento机构：Indian Institute of Technology Kharagpur, India, Queen Mary University of London, UK摘要：子词标记化在基于Transformer模型的基于文本的自然语言处理（NLP）任务中取得了广泛的成功。随着Transformer模型在符号音乐相关研究中越来越受欢迎，研究符号音乐域中的子词标记化的功效势在必行。在本文中，我们探讨了子词标记化技术，如字节对编码（BPE），在符号音乐生成及其对生成的歌曲的整体结构的影响。我们的实验基于三种类型的MIDI数据集：仅单音轨旋律、具有单个乐器的多音轨、以及多音轨和多乐器。我们应用子词标记后的音乐标记化方案，发现它能够生成较长的歌曲在同一时间，并提高了整体结构的客观指标，如结构指标（SI），音高类熵等方面所生成的音乐，我们还比较了两个子词标记化方法，BPE和Unigram，并观察到这两种方法导致一致的改进。我们的研究表明，子词标记化是一种很有前途的符号音乐生成技术，并可能对音乐创作产生更广泛的影响，特别是在涉及复杂数据的情况下，如多轨歌曲。摘要：Subword tokenization has been widely successful in text-based natural language processing (NLP) tasks with Transformer-based models. As Transformer models become increasingly popular in symbolic music-related studies, it is imperative to investigate the efficacy of subword tokenization in the symbolic music domain. In this paper, we explore subword tokenization techniques, such as byte-pair encoding (BPE), in symbolic music generation and its impact on the overall structure of generated songs. Our experiments are based on three types of MIDI datasets: single track-melody only, multi-track with a single instrument, and multi-track and multi-instrument. We apply subword tokenization on post-musical tokenization schemes and find that it enables the generation of longer songs at the same time and improves the overall structure of the generated music in terms of objective metrics like structure indicator (SI), Pitch Class Entropy, etc. We also compare two subword tokenization methods, BPE and Unigram, and observe that both methods lead to consistent improvements. Our study suggests that subword tokenization is a promising technique for symbolic music generation and may have broader implications for music composition, particularly in cases involving complex data such as multi-track songs.

【7】Approximate Nearest Neighbour Phrase Mining for Contextual Speech Recognition标题：上下文语音识别中的近似最近邻短语挖掘链接：http://arxiv.org/pdf/2304.08862v1

作者：Maurits Bleeker,Pawel Swietojanski,Stefan Braun,Xiaodan Zhuang机构：University of Amsterdam, The Netherlands, Apple备注：5 pages, 2 figures, 2 tables摘要：本文提出了一种扩展训练端到端的上下文感知Transformer器传感器（CATT）模型，通过使用一个简单的，但有效的方法挖掘硬否定短语的潜在空间的上下文编码器。在训练过程中，给定一个参考查询，我们使用近似最近邻搜索挖掘一些相似的短语。然后，这些采样的短语被用作上下文列表中的否定示例，以及随机和地面实况上下文信息。通过在上下文列表中包括近似最近邻短语（ANN-P），我们鼓励学习的表示在相似但不相同的偏置短语之间消除歧义。当偏置库存中存在若干类似短语时，这提高了偏置准确性。我们进行实验，在大规模的数据制度获得高达7%的相对字错误率降低测试数据的上下文部分。我们还扩展和评估CATT方法在流媒体应用程序。摘要：This paper presents an extension to train end-to-end Context-Aware Transformer Transducer ( CATT ) models by using a simple, yet efficient method of mining hard negative phrases from the latent space of the context encoder. During training, given a reference query, we mine a number of similar phrases using approximate nearest neighbour search. These sampled phrases are then used as negative examples in the context list alongside random and ground truth contextual information. By including approximate nearest neighbour phrases (ANN-P) in the context list, we encourage the learned representation to disambiguate between similar, but not identical, biasing phrases. This improves biasing accuracy when there are several similar phrases in the biasing inventory. We carry out experiments in a large-scale data regime obtaining up to 7% relative word error rate reductions for the contextual portion of test data. We also extend and evaluate CATT approach in streaming applications.

【8】Towards the Transferable Audio Adversarial Attack via Ensemble Methods标题：基于集成方法的可转移音频对抗攻击链接：http://arxiv.org/pdf/2304.08811v1

*与cs.SD语音【3】为同一篇

作者：Feng Guo,Zheng Sun,Yuxuan Chen,Lei Ju机构：cn 1School of Cyber Science and Technology, Shandong University备注：Submitted to Cybersecurity journal 2023摘要：近年来，深度学习（DL）模型在许多领域取得了重大进展，例如自动驾驶，面部识别和语音识别。然而，深度学习模型对对抗性攻击的脆弱性引起了社区的严重关注，因为它们的鲁棒性和泛化能力不足。此外，可转移攻击已经成为黑盒攻击的一种重要方法。在这项工作中，我们探讨了影响基于DL的语音识别中对抗性示例（AE）可转移性的潜在因素。我们还讨论了不同的DL系统的脆弱性和决策边界的不规则性。我们的研究结果表明，语音和图像之间的AE的可转移性显着的差异，与数据的相关性低的图像，但在语音识别相反。出于辍学为基础的集成方法，我们提出了随机梯度合奏和动态梯度加权合奏，我们评估合奏的影响AE的可转移性。结果表明，由这两种方法创建的AE对于转移到黑盒API是有效的。摘要：In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufficient robustness and generalization. Also, transferable attacks have become a prominent method for black-box attacks. In this work, we explore the potential factors that impact adversarial examples (AEs) transferability in DL-based speech recognition. We also discuss the vulnerability of different DL systems and the irregular nature of decision boundaries. Our results show a remarkable difference in the transferability of AEs between speech and images, with the data relevance being low in images but opposite in speech recognition. Motivated by dropout-based ensemble approaches, we propose random gradient ensembles and dynamic gradient-weighted ensembles, and we evaluate the impact of ensembles on the transferability of AEs. The results show that the AEs created by both approaches are valid for transfer to the black box API.

【9】A Voice Disease Detection Method Based on MFCCs and Shallow CNN标题：一种基于MFCCS和浅层CNN的语音疾病检测方法链接：http://arxiv.org/pdf/2304.08708v1

*与cs.SD语音【4】为同一篇

作者：Xiaoping Xie,Hao Cai,Can Li,Fei Ding机构：We cooperated with Xiangya Hospital of Central South University to collect voice samples from sixty-one different patients, Author was with Rice University, He is now with the Department of Physics, Colorado State University摘要：嗓音疾病的发病率呈逐年上升趋势。利用软件进行远程诊断是一种技术发展趋势，具有重要的实用价值。在嗓音疾病中，引起声音嘶哑的常见疾病包括痉挛性发声障碍、声带麻痹、声带小结和声带息肉。本文提出了一种可广泛应用于临床的嗓音疾病检测方法。我们与中南大学湘雅医院合作，收集了61名不同患者的声音样本。提取Mel频率倒谱系数（MFCC）参数作为输入特征，以数据的形式描述语音。提出了一种结合MFCC参数和单卷积层CNN的创新模型，用于快速计算和分类。我们达到的最高准确率为92%，这是完全领先于原来的研究成果和国际先进水平。并利用高级语音功能评估数据库（AVFAD）对该方法的泛化能力进行了测试，准确率达到98%。在临床和标准数据集上的实验表明，对于嗓音疾病的病理检测，该方法在准确率和计算效率上都有很大的提高。摘要：The incidence rate of voice diseases is increasing year by year. The use of software for remote diagnosis is a technical development trend and has important practical value. Among voice diseases, common diseases that cause hoarseness include spasmodic dysphonia, vocal cord paralysis, vocal nodule, and vocal cord polyp. This paper presents a voice disease detection method that can be applied in a wide range of clinical. We cooperated with Xiangya Hospital of Central South University to collect voice samples from sixty-one different patients. The Mel Frequency Cepstrum Coefficient (MFCC) parameters are extracted as input features to describe the voice in the form of data. An innovative model combining MFCC parameters and single convolution layer CNN is proposed for fast calculation and classification. The highest accuracy we achieved was 92%, it is fully ahead of the original research results and internationally advanced. And we use Advanced Voice Function Assessment Databases (AVFAD) to evaluate the generalization ability of the method we proposed, which achieved an accuracy rate of 98%. Experiments on clinical and standard datasets show that for the pathological detection of voice diseases, our method has greatly improved in accuracy and computational efficiency.

【10】Evolution of the Informational Complexity of Contemporary Western Music标题：当代西方音乐信息复杂性的演变链接：http://arxiv.org/pdf/1907.04292v2

作者：Thomas Parmer,Yong-Yeol Ahn机构：School of Informatics, Computing, and Engineering, Indiana University, Bloomington, IN, USA备注：8 pages, 6 figures; added supplementary materials摘要：我们衡量的歌曲的复杂性在百万歌曲数据集（MSD）的音高，音色，响度和节奏，调查他们的演变从1960年到2010年。通过比较公告牌热门100与随机样本，我们发现，流行歌曲的复杂性往往是更窄的平均值周围分布，支持复杂性和享乐价值之间的倒U形关系的想法。然后，我们研究的时间演变的复杂性，报告一致的变化跨越几十年，如自20世纪60年代以来的平均响度复杂性的下降，并增加音色复杂性的整体，但不是流行歌曲。我们还表明，与流行歌曲随着时间的推移听起来更相似的说法相反，它们在音高或节奏方面并不比50年前更相似，尽管音色的相似性显示了不同时代的独特模式，并且响度的相似性一直在增加。最后，我们表明，音乐流派可以区分其独特的复杂性配置文件。摘要：We measure the complexity of songs in the Million Song Dataset (MSD) in terms of pitch, timbre, loudness, and rhythm to investigate their evolution from 1960 to 2010. By comparing the Billboard Hot 100 with random samples, we find that the complexity of popular songs tends to be more narrowly distributed around the mean, supporting the idea of an inverted U-shaped relationship between complexity and hedonistic value. We then examine the temporal evolution of complexity, reporting consistent changes across decades, such as a decrease in average loudness complexity since the 1960s, and an increase in timbre complexity overall but not for popular songs. We also show, in contrast to claims that popular songs sound more alike over time, that they are not more similar than they were 50 years ago in terms of pitch or rhythm, although similarity in timbre shows distinctive patterns across eras and similarity in loudness has been increasing. Finally, we show that musical genres can be differentiated by their distinctive complexity profiles.

机器翻译，仅供参考