今日论文合集：cs.SD语音8篇，eess.AS音频处理9篇。
cs.SD语音
【1】 Machine Learning-based Classification of Birds through Birdsong
标题：基于机器学习的鸟鸣鸟类分类
链接：https://arxiv.org/abs/2212.04684
作者：Yueying Chang,Richard O. Sinnott机构：The University of Melbourne, Australia摘要：Audio sound recognition and classification is used for many tasks and applications including human voice recognition, music recognition and audio tagging. In this paper we apply Mel Frequency Cepstral Coefficients (MFCC) in combination with a range of machine learning models to identify (Australian) birds from publicly available audio files of their birdsong. We present approaches used for data processing and augmentation and compare the results of various state of the art machine learning models. We achieve an overall accuracy of 91% for the top-5 birds from the 30 selected as the case study. Applying the models to more challenging and diverse audio files comprising 152 bird species, we achieve an accuracy of 58%
【2】 Hyperbolic Audio Source Separation
标题：双曲音频源分离
链接：https://arxiv.org/abs/2212.05008
作者：Darius Petermann,Gordon Wichern,Aswin Subramanian,Jonathan Le Roux机构：Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA, USA, Indiana University, Department of Intelligent Systems Engineering, Bloomington, IN, USA备注：Submitted to ICASSP 2023, Demo page: this https URL摘要：We introduce a framework for audio source separation using embeddings on a hyperbolic manifold that compactly represent the hierarchical relationship between sound sources and time-frequency features. Inspired by recent successes modeling hierarchical relationships in text and images with hyperbolic embeddings, our algorithm obtains a hyperbolic embedding for each time-frequency bin of a mixture signal and estimates masks using hyperbolic softmax layers. On a synthetic dataset containing mixtures of multiple people talking and musical instruments playing, our hyperbolic model performed comparably to a Euclidean baseline in terms of source to distortion ratio, with stronger performance at low embedding dimensions. Furthermore, we find that time-frequency regions containing multiple overlapping sources are embedded towards the center (i.e., the most uncertain region) of the hyperbolic space, and we can use this certainty estimate to efficiently trade-off between artifact introduction and interference reduction when isolating individual sounds.
【3】 DDSupport: Language Learning Support System that Displays Differences and Distances from Model Speech
标题：DDSupport：语言学习支持系统，显示与模拟语音的差异和距离
链接：https://arxiv.org/abs/2212.04930
作者：Kazuki Kawamura,Jun Rekimoto机构：The University of Tokyo, Tokyo, Japan, Sony CSL Kyoto, Kyoto, Japan摘要：当初学者学习说一种非母语的语言时，他们很难自己判断自己说得是否好。因此，使用计算机辅助发音训练系统来检测学习者的错误发音。这些系统通常以节奏、音素或单词为单位将用户的语音与作为模型的特定母语说话者的语音进行比较，并计算差异。然而，他们需要大量的语音数据和详细的注释，或者只能与一个特定的母语使用者进行比较。为了克服这些问题，我们提出了一种新的语言学习支持系统，其基于少量未注释的语音数据计算语音得分并检测初学者的错误发音，而不与特定的人进行比较。所提出的系统使用基于深度学习的语音处理来以直观可视的方式显示学习者的语音的发音得分以及学习者的发音与一组模型的发音之间的差异/距离。学习者可以通过消除差异和缩短与模型的距离来逐渐改善他们的发音，直到他们足够熟练。此外，由于与特定模型的特定句子相比，发音得分和差异/距离没有被计算，因此用户可以自由地学习他们希望学习的句子。我们还开发了一个应用程序来帮助非英语母语的人学习英语，并证实它可以提高用户的语音清晰度。摘要：When beginners learn to speak a non-native language, it is difficult for them to judge for themselves whether they are speaking well. Therefore, computer-assisted pronunciation training systems are used to detect learner mispronunciations. These systems typically compare the user's speech with that of a specific native speaker as a model in units of rhythm, phonemes, or words and calculate the differences. However, they require extensive speech data with detailed annotations or can only compare with one specific native speaker. To overcome these problems, we propose a new language learning support system that calculates speech scores and detects mispronunciations by beginners based on a small amount of unannotated speech data without comparison to a specific person. The proposed system uses deep learning--based speech processing to display the pronunciation score of the learner's speech and the difference/distance between the learner's and a group of models' pronunciation in an intuitively visual manner. Learners can gradually improve their pronunciation by eliminating differences and shortening the distance from the model until they become sufficiently proficient. Furthermore, since the pronunciation score and difference/distance are not calculated compared to specific sentences of a particular model, users are free to study the sentences they wish to study. We also built an application to help non-native speakers learn English and confirmed that it can improve users' speech intelligibility.
【4】 Uncertainty Estimation in Deep Speech Enhancement Using Complex Gaussian Mixture Models
标题：基于复高斯混合模型的深度语音增强不确定性估计
链接：https://arxiv.org/abs/2212.04831
作者：Huajian Fang,Timo Gerkmann机构：Signal Processing (SP), Universit¨at Hamburg, Germany摘要：单通道深度语音增强方法通常估计单个乘法掩模来提取干净语音，而不测量其准确度。相反，在这项工作中，我们建议量化与基于神经网络的语音增强中的干净语音估计相关的不确定性。预测不确定性通常分为随机不确定性和认识不确定性。前者说明数据中固有的不确定性，后者对应于模型不确定性。为了实现鲁棒的干净语音估计和有效的预测不确定性量化，提出将统计复高斯混合模型（CGMM）集成到深度语音增强框架中。更具体地，我们通过条件概率密度随机地对输入和输出之间的依赖性进行建模，并且训练神经网络以将噪声输入映射到干净语音的完全后验分布，干净语音被建模为多个复高斯分量的混合。在不同数据集上的实验结果表明，该算法有效地捕捉了预测的不确定性，结合强大的统计模型和深度学习也能提供更好的语音增强性能。摘要：Single-channel deep speech enhancement approaches often estimate a single multiplicative mask to extract clean speech without a measure of its accuracy. Instead, in this work, we propose to quantify the uncertainty associated with clean speech estimates in neural network-based speech enhancement. Predictive uncertainty is typically categorized into aleatoric uncertainty and epistemic uncertainty. The former accounts for the inherent uncertainty in data and the latter corresponds to the model uncertainty. Aiming for robust clean speech estimation and efficient predictive uncertainty quantification, we propose to integrate statistical complex Gaussian mixture models (CGMMs) into a deep speech enhancement framework. More specifically, we model the dependency between input and output stochastically by means of a conditional probability density and train a neural network to map the noisy input to the full posterior distribution of clean speech, modeled as a mixture of multiple complex Gaussian components. Experimental results on different datasets show that the proposed algorithm effectively captures predictive uncertainty and that combining powerful statistical models and deep learning also delivers a superior speech enhancement performance.
【5】 High Quality Audio Coding with MDCTNet
标题：用MDCTNet实现高质量的音频编码
链接：https://arxiv.org/abs/2212.04583
作者：Grant Davidson,Mark Vinton,Per Ekstrand,Cong Zhou,Lars Villemoes,Lie Lu机构：Dolby Laboratories, Inc., Market Street, San Francisco, CA, USA, Dolby Sweden AB, Gävlegatan ,A, Stockholm, Sweden备注：Five pages, five figures摘要：提出了一种在自适应修正离散余弦变换（MDCT）的感知加权域中工作的神经音频生成模型MDCTNet。该模型的结构利用递归层（RNN）捕获时间和频率方向上的相关性。通过在由感知音频编码器调节的48 kHz采样的全频带单声道音频信号的不同集合上训练MDCTNet来获得音频编码系统。在具有被选择为在内容类型之间平衡的10个摘录的主观收听测试中，然而对于两个编解码器都有压力，对于24 kb/s可变比特率（VBR），所提出的系统的平均性能类似于Opus在两倍比特率时的平均性能。摘要：We propose a neural audio generative model, MDCTNet, operating in the perceptually weighted domain of an adaptive modified discrete cosine transform (MDCT). The architecture of the model captures correlations in both time and frequency directions with recurrent layers (RNNs). An audio coding system is obtained by training MDCTNet on a diverse set of fullband monophonic audio signals at 48 kHz sampling, conditioned by a perceptual audio encoder. In a subjective listening test with ten excerpts chosen to be balanced across content types, yet stressful for both codecs, the mean performance of the proposed system for 24 kb/s variable bitrate (VBR) is similar to that of Opus at twice the bitrate.
【6】 A Data-driven Cognitive Salience Model for Objective Perceptual Audio Quality Assessment
标题：用于客观感知音频质量评估的数据驱动认知显著模型
链接：https://arxiv.org/abs/2212.04572
作者：Pablo M. Delgado,Jürgen Herre机构：⋆ International Audio Laboratories Erlangen ‡, Am Wolfsmantel , Erlangen, Germany, † Fraunhofer IIS, Am Wolfsmantel , Erlangen, Germany摘要：客观音频质量测量系统通常使用感知模型来预测经处理的信号的主观质量得分，如在收听测试中所报告的。大多数系统将感知到的退化的不同度量映射到预测主观质量的单个质量分数。这需要由使用统计学习的真实收听测试数据通知的质量映射阶段（即，数据驱动方法），其中失真度量作为输入特征。然而，在实践中，可靠的训练数据量是有限的，并且通常不足以对大型学习模型进行全面训练。然而，客观系统中的认知效应模型可以改进学习模型。具体而言，考虑到某些失真类型的显著性，它们向映射阶段提供了改进学习过程的附加特征，尤其是对于有限量的训练数据。我们提出了一种新的数据驱动的显著性模型，该模型通过使用显著性度量显式地估计认知/退化度量的交互作用来通知质量映射阶段。对于代表性的验证数据集，并入了新颖的显著性模型的系统被示出优于仅使用统计学习来组合认知和退化度量的等效系统以及其他公知的测量系统。摘要：Objective audio quality measurement systems often use perceptual models to predict the subjective quality scores of processed signals, as reported in listening tests. Most systems map different metrics of perceived degradation into a single quality score predicting subjective quality. This requires a quality mapping stage that is informed by real listening test data using statistical learning (i.e., a data-driven approach) with distortion metrics as input features. However, the amount of reliable training data is limited in practice, and usually not sufficient for a comprehensive training of large learning models. Models of cognitive effects in objective systems can, however, improve the learning model. Specifically, considering the salience of certain distortion types, they provide additional features to the mapping stage that improve the learning process, especially for limited amounts of training data. We propose a novel data-driven salience model that informs the quality mapping stage by explicitly estimating the cognitive/degradation metric interactions using a salience measure. Systems incorporating the novel salience model are shown to outperform equivalent systems that only use statistical learning to combine cognitive and degradation metrics, as well as other well-known measurement systems, for a representative validation dataset.
【7】 SpeechLMScore: Evaluating speech generation using speech language model
标题：SpeechLMScore：使用语音语言模型评估语音生成
链接：https://arxiv.org/abs/2212.04559
作者：Soumi Maiti,Yifan Peng,Takaaki Saeki,Shinji Watanabe机构：Carnegie Mellon University, USA,The University of Tokyo, Japan摘要：While human evaluation is the most reliable metric for evaluating speech generation systems, it is generally costly and time-consuming. Previous studies on automatic speech quality assessment address the problem by predicting human evaluation scores with machine learning models. However, they rely on supervised learning and thus suffer from high annotation costs and domain-shift problems. We propose SpeechLMScore, an unsupervised metric to evaluate generated speech using a speech-language model. SpeechLMScore computes the average log-probability of a speech signal by mapping it into discrete tokens and measures the average probability of generating the sequence of tokens. Therefore, it does not require human annotation and is a highly scalable framework. Evaluation results demonstrate that the proposed metric shows a promising correlation with human evaluation scores on different speech generation tasks including voice conversion, text-to-speech, and speech enhancement.
【8】 Framewise WaveGAN: High Speed Adversarial Vocoder in Time Domain with Very Low Computational Complexity
标题：FRAMEWAGE WaveGAN：计算复杂度极低的高速对抗性时间域声码器
链接：https://arxiv.org/abs/2212.04532
作者：Ahmed Mustafa,Jean-Marc Valin,Jan Büthe,Paris Smaragdis,Mike Goodwin机构：⋆Amazon Web Services, Palo Alto, USA, †University of Illinois at Urbana-Champaign备注：Submitted to ICASSP 2023, demo: this https URL摘要：GAN声码器是目前用于建立高质量神经波形生成模型的最先进的方法之一。然而，它们的大多数架构需要每秒几十亿次浮点运算（GFLOPS）来以采样方式生成语音波形。这使得GAN声码器在没有加速器或并行计算机的普通CPU上运行仍然具有挑战性。本文提出了一种新的GAN声码器结构，主要依靠递归全连通网络直接产生时域信号。这导致计算成本的显著降低，并且使得能够在GPU和低复杂性CPU上非常快速地生成。实验结果表明，我们的Framewise WaveGAN声码器以1.2GFLOPS的非常低的复杂度实现了比诸如LPCNet的自回归最大似然声码器显著更高的质量。这使得GAN声码器在边缘和低功耗设备上更加实用。摘要：GAN vocoders are currently one of the state-of-the-art methods for building high-quality neural waveform generative models. However, most of their architectures require dozens of billion floating-point operations per second (GFLOPS) to generate speech waveforms in samplewise manner. This makes GAN vocoders still challenging to run on normal CPUs without accelerators or parallel computers. In this work, we propose a new architecture for GAN vocoders that mainly depends on recurrent and fully-connected networks to directly generate the time domain signal in framewise manner. This results in considerable reduction of the computational cost and enables very fast generation on both GPUs and low-complexity CPUs. Experimental results show that our Framewise WaveGAN vocoder achieves significantly higher quality than auto-regressive maximum-likelihood vocoders such as LPCNet at a very low complexity of 1.2 GFLOPS. This makes GAN vocoders more practical on edge and low-power devices.
【1】 Hyperbolic Audio Source Separation
标题：双曲音频源分离
链接：https://arxiv.org/abs/2212.05008
*与cs.SD语音【2】为同一篇
作者：Darius Petermann,Gordon Wichern,Aswin Subramanian,Jonathan Le Roux机构：Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA, USA, Indiana University, Department of Intelligent Systems Engineering, Bloomington, IN, USA备注：Submitted to ICASSP 2023, Demo page: this https URL摘要：We introduce a framework for audio source separation using embeddings on a hyperbolic manifold that compactly represent the hierarchical relationship between sound sources and time-frequency features. Inspired by recent successes modeling hierarchical relationships in text and images with hyperbolic embeddings, our algorithm obtains a hyperbolic embedding for each time-frequency bin of a mixture signal and estimates masks using hyperbolic softmax layers. On a synthetic dataset containing mixtures of multiple people talking and musical instruments playing, our hyperbolic model performed comparably to a Euclidean baseline in terms of source to distortion ratio, with stronger performance at low embedding dimensions. Furthermore, we find that time-frequency regions containing multiple overlapping sources are embedded towards the center (i.e., the most uncertain region) of the hyperbolic space, and we can use this certainty estimate to efficiently trade-off between artifact introduction and interference reduction when isolating individual sounds.
【2】 DDSupport: Language Learning Support System that Displays Differences and Distances from Model Speech
标题：DDSupport：语言学习支持系统，显示与模拟语音的差异和距离
链接：https://arxiv.org/abs/2212.04930
*与cs.SD语音【3】为同一篇
作者：Kazuki Kawamura,Jun Rekimoto机构：The University of Tokyo, Tokyo, Japan, Sony CSL Kyoto, Kyoto, Japan摘要：When beginners learn to speak a non-native language, it is difficult for them to judge for themselves whether they are speaking well. Therefore, computer-assisted pronunciation training systems are used to detect learner mispronunciations. These systems typically compare the user's speech with that of a specific native speaker as a model in units of rhythm, phonemes, or words and calculate the differences. However, they require extensive speech data with detailed annotations or can only compare with one specific native speaker. To overcome these problems, we propose a new language learning support system that calculates speech scores and detects mispronunciations by beginners based on a small amount of unannotated speech data without comparison to a specific person. The proposed system uses deep learning--based speech processing to display the pronunciation score of the learner's speech and the difference/distance between the learner's and a group of models' pronunciation in an intuitively visual manner. Learners can gradually improve their pronunciation by eliminating differences and shortening the distance from the model until they become sufficiently proficient. Furthermore, since the pronunciation score and difference/distance are not calculated compared to specific sentences of a particular model, users are free to study the sentences they wish to study. We also built an application to help non-native speakers learn English and confirmed that it can improve users' speech intelligibility.
【3】 Uncertainty Estimation in Deep Speech Enhancement Using Complex Gaussian Mixture Models
标题：基于复高斯混合模型的深度语音增强不确定性估计
链接：https://arxiv.org/abs/2212.04831
*与cs.SD语音【4】为同一篇
作者：Huajian Fang,Timo Gerkmann机构：Signal Processing (SP), Universit¨at Hamburg, Germany摘要：Single-channel deep speech enhancement approaches often estimate a single multiplicative mask to extract clean speech without a measure of its accuracy. Instead, in this work, we propose to quantify the uncertainty associated with clean speech estimates in neural network-based speech enhancement. Predictive uncertainty is typically categorized into aleatoric uncertainty and epistemic uncertainty. The former accounts for the inherent uncertainty in data and the latter corresponds to the model uncertainty. Aiming for robust clean speech estimation and efficient predictive uncertainty quantification, we propose to integrate statistical complex Gaussian mixture models (CGMMs) into a deep speech enhancement framework. More specifically, we model the dependency between input and output stochastically by means of a conditional probability density and train a neural network to map the noisy input to the full posterior distribution of clean speech, modeled as a mixture of multiple complex Gaussian components. Experimental results on different datasets show that the proposed algorithm effectively captures predictive uncertainty and that combining powerful statistical models and deep learning also delivers a superior speech enhancement performance.
【4】 Geometry-aware DoA Estimation using a Deep Neural Network with mixed-data input features
标题：基于混合数据输入特征的深度神经网络几何感知波达方向估计
链接：https://arxiv.org/abs/2212.04788
作者：Ulrik Kowalk,Simon Doclo,Joerg Bitzer机构：Jade University of Applied Sciences, Institute for Hearing Technology and Audiology, Oldenburg, Germany, University of Oldenburg, Department of Medical Physics and Acoustics and, Cluster of Excellence Hearing,all, Oldenburg, Germany备注：Submitted to ICASSP 2023摘要：Unlike model-based direction of arrival (DoA) estimation algorithms, supervised learning-based DoA estimation algorithms based on deep neural networks (DNNs) are usually trained for one specific microphone array geometry, resulting in poor performance when applied to a different array geometry. In this paper we illustrate the fundamental difference between supervised learning-based and model-based algorithms leading to this sensitivity. Aiming at designing a supervised learning-based DoA estimation algorithm that generalizes well to different array geometries, in this paper we propose a geometry-aware DoA estimation algorithm. The algorithm uses a fully connected DNN and takes mixed data as input features, namely the time lags maximizing the generalized cross-correlation with phase transform and the microphone coordinates, which are assumed to be known. Experimental results for a reverberant scenario demonstrate the flexibility of the proposed algorithm towards different array geometries and show that the proposed algorithm outperforms model-based algorithms such as steered response power with phase transform.
【5】 High Quality Audio Coding with MDCTNet
标题：用MDCTNet实现高质量的音频编码
链接：https://arxiv.org/abs/2212.04583
*与cs.SD语音【5】为同一篇
作者：Grant Davidson,Mark Vinton,Per Ekstrand,Cong Zhou,Lars Villemoes,Lie Lu机构：Dolby Laboratories, Inc., Market Street, San Francisco, CA, USA, Dolby Sweden AB, Gävlegatan ,A, Stockholm, Sweden备注：Five pages, five figures摘要：We propose a neural audio generative model, MDCTNet, operating in the perceptually weighted domain of an adaptive modified discrete cosine transform (MDCT). The architecture of the model captures correlations in both time and frequency directions with recurrent layers (RNNs). An audio coding system is obtained by training MDCTNet on a diverse set of fullband monophonic audio signals at 48 kHz sampling, conditioned by a perceptual audio encoder. In a subjective listening test with ten excerpts chosen to be balanced across content types, yet stressful for both codecs, the mean performance of the proposed system for 24 kb/s variable bitrate (VBR) is similar to that of Opus at twice the bitrate.
【6】 A Data-driven Cognitive Salience Model for Objective Perceptual Audio Quality Assessment
标题：用于客观感知音频质量评估的数据驱动认知显著模型
链接：https://arxiv.org/abs/2212.04572
*与cs.SD语音【6】为同一篇
作者：Pablo M. Delgado,Jürgen Herre机构：⋆ International Audio Laboratories Erlangen ‡, Am Wolfsmantel , Erlangen, Germany, † Fraunhofer IIS, Am Wolfsmantel , Erlangen, Germany摘要：客观音频质量测量系统通常使用感知模型来预测经处理的信号的主观质量得分，如在收听测试中所报告的。大多数系统将感知到的退化的不同度量映射到预测主观质量的单个质量分数。这需要由使用统计学习的真实收听测试数据通知的质量映射阶段（即，数据驱动方法），其中失真度量作为输入特征。然而，在实践中，可靠的训练数据量是有限的，并且通常不足以对大型学习模型进行全面训练。然而，客观系统中的认知效应模型可以改进学习模型。具体而言，考虑到某些失真类型的显著性，它们向映射阶段提供了改进学习过程的附加特征，尤其是对于有限量的训练数据。我们提出了一种新的数据驱动的显著性模型，通过使用显著性度量显式地估计认知/退化度量的交互作用来通知质量映射阶段。对于代表性的验证数据集，并入了新颖的显著性模型的系统被示出优于仅使用统计学习来组合认知和退化度量的等效系统以及其他公知的测量系统。摘要：Objective audio quality measurement systems often use perceptual models to predict the subjective quality scores of processed signals, as reported in listening tests. Most systems map different metrics of perceived degradation into a single quality score predicting subjective quality. This requires a quality mapping stage that is informed by real listening test data using statistical learning (i.e., a data-driven approach) with distortion metrics as input features. However, the amount of reliable training data is limited in practice, and usually not sufficient for a comprehensive training of large learning models. Models of cognitive effects in objective systems can, however, improve the learning model. Specifically, considering the salience of certain distortion types, they provide additional features to the mapping stage that improve the learning process, especially for limited amounts of training data. We propose a novel data-driven salience model that informs the quality mapping stage by explicitly estimating the cognitive/degradation metric interactions using a salience measure. Systems incorporating the novel salience model are shown to outperform equivalent systems that only use statistical learning to combine cognitive and degradation metrics, as well as other well-known measurement systems, for a representative validation dataset.
【7】 SpeechLMScore: Evaluating speech generation using speech language model
标题：SpeechLMScore：使用语音语言模型评估语音生成
链接：https://arxiv.org/abs/2212.04559
*与cs.SD语音【7】为同一篇
作者：Soumi Maiti,Yifan Peng,Takaaki Saeki,Shinji Watanabe机构：Carnegie Mellon University, USA,The University of Tokyo, Japan摘要：While human evaluation is the most reliable metric for evaluating speech generation systems, it is generally costly and time-consuming. Previous studies on automatic speech quality assessment address the problem by predicting human evaluation scores with machine learning models. However, they rely on supervised learning and thus suffer from high annotation costs and domain-shift problems. We propose SpeechLMScore, an unsupervised metric to evaluate generated speech using a speech-language model. SpeechLMScore computes the average log-probability of a speech signal by mapping it into discrete tokens and measures the average probability of generating the sequence of tokens. Therefore, it does not require human annotation and is a highly scalable framework. Evaluation results demonstrate that the proposed metric shows a promising correlation with human evaluation scores on different speech generation tasks including voice conversion, text-to-speech, and speech enhancement.
【8】 Framewise WaveGAN: High Speed Adversarial Vocoder in Time Domain with Very Low Computational Complexity
标题：FRAMEWAGE WaveGAN：计算复杂度极低的高速对抗性时间域声码器
链接：https://arxiv.org/abs/2212.04532
*与cs.SD语音【8】为同一篇
作者：Ahmed Mustafa,Jean-Marc Valin,Jan Büthe,Paris Smaragdis,Mike Goodwin机构：⋆Amazon Web Services, Palo Alto, USA, †University of Illinois at Urbana-Champaign备注：Submitted to ICASSP 2023, demo: this https URL摘要：GAN vocoders are currently one of the state-of-the-art methods for building high-quality neural waveform generative models. However, most of their architectures require dozens of billion floating-point operations per second (GFLOPS) to generate speech waveforms in samplewise manner. This makes GAN vocoders still challenging to run on normal CPUs without accelerators or parallel computers. In this work, we propose a new architecture for GAN vocoders that mainly depends on recurrent and fully-connected networks to directly generate the time domain signal in framewise manner. This results in considerable reduction of the computational cost and enables very fast generation on both GPUs and low-complexity CPUs. Experimental results show that our Framewise WaveGAN vocoder achieves significantly higher quality than auto-regressive maximum-likelihood vocoders such as LPCNet at a very low complexity of 1.2 GFLOPS. This makes GAN vocoders more practical on edge and low-power devices.
【9】 Machine Learning-based Classification of Birds through Birdsong
标题：基于机器学习的鸟鸣鸟类分类
链接：https://arxiv.org/abs/2212.04684
*与cs.SD语音【1】为同一篇
作者：Yueying Chang,Richard O. Sinnott机构：The University of Melbourne, Australia摘要：Audio sound recognition and classification is used for many tasks and applications including human voice recognition, music recognition and audio tagging. In this paper we apply Mel Frequency Cepstral Coefficients (MFCC) in combination with a range of machine learning models to identify (Australian) birds from publicly available audio files of their birdsong. We present approaches used for data processing and augmentation and compare the results of various state of the art machine learning models. We achieve an overall accuracy of 91% for the top-5 birds from the 30 selected as the case study. Applying the models to more challenging and diverse audio files comprising 152 bird species, we achieve an accuracy of 58%
机器翻译，仅供参考
每日语音速递