Towards Robust Neural Vocoding for Speech Generation: A Survey
Towards Robust Neural Vocoding for Speech Generation: A Survey
复制标题
用于语音生成的鲁棒神经声码：一项调查
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Hung
中科院分区：
文献类型：
--
作者：
Po;Chun;Andy T. Liu;Hung
最近，神经声码器已广泛应用于语音合成任务，包括文本到语音和语音转换。然而，当遇到训练和推理之间的数据分布不匹配时，在真实数据上训练的神经声码器通常会在未见过的场景中降低语音质量。在本文中，我们在五个不同的数据集上交替训练四种常见的神经声码器，包括 WaveNet、WaveRNN、FFTNet、Parallel WaveGAN。为了研究神经声码器的鲁棒性，我们使用来自见过/没见过的说话者、见过/没见过的语言、文本到语音模型和语音转换模型的声学特征来评估模型。我们发现，对于实现通用声码器而言，说话者的多样性比语言更为重要。通过我们的实验，我们表明 WaveNet 和 WaveRNN 更适合文本转语音模型，而 Parallel WaveGAN 更适合语音转换应用。所有声码器自然度方面的大量主观 MOS 结果可供未来研究使用。
Recently, neural vocoders have been widely used in speech synthesis tasks, including text-to-speech and voice conversion. However, when encountering data distribution mismatch between training and inference, neural vocoders trained on real data often degrade in voice quality for unseen scenarios. In this paper, we train four common neural vocoders, including WaveNet, WaveRNN, FFTNet, Parallel WaveGAN alternately on five different datasets. To study the robustness of neural vocoders, we evaluate the models using acoustic features from seen/unseen speakers, seen/unseen languages, a text-to-speech model, and a voice conversion model. We found out that the speaker variety is much more important for achieving a universal vocoder than the language. Through our experiments, we show that WaveNet and WaveRNN are more suitable for text-to-speech models, while Parallel WaveGAN is more suitable for voice conversion applications. Great amount of subjective MOS results in naturalness for all vocoders are presented for future studies.