End-to-end text-to-speech synthesis (TTS) can generate highly natural synthetic speech from raw text. However, rendering the correct pitch accents is still a challenging problem for end-to-end TTS. To tackle the challenge of rendering correct pitch accent in Japanese end-to-end TTS, we adopt PnG~BERT, a self-supervised pretrained model in the character and phoneme domain for TTS. We investigate the effects of features captured by PnG~BERT on Japanese TTS by modifying the fine-tuning condition to determine the conditions helpful inferring pitch accents. We manipulate content of PnG~BERT features from being text-oriented to speech-oriented by changing the number of fine-tuned layers during TTS. In addition, we teach PnG~BERT pitch accent information by fine-tuning with tone prediction as an additional downstream task. Our experimental results show that the features of PnG~BERT captured by pretraining contain information helpful inferring pitch accent, and PnG~BERT outperforms baseline Tacotron on accent correctness in a listening test.
The criteria for measuring music similarity are important for developing a flexible music recommendation system. Some data-driven methods have been proposed to calculate music similarity from only music signals, such as metric learning based on a triplet loss using tag information on each musical piece. However, the resulting music similarity metric usually captures the entire piece of music, i.e., the mixing of various instrumental sound sources, limiting the capability of the music recommendation system, e.g., it is difficult to search for a musical piece containing similar drum sounds. Towards the development of a more flexible music recommendation system, we propose a music similarity calculation method that focuses on individual instrumental sound sources in a musical piece. By fully exploiting the potential of data-driven methods for our proposed method, we employ weakly supervised metric learning to individual instrumental sound source signals without using any tag information, where positive and negative samples in a triplet loss are defined by whether or not they are from the same musical piece. Furthermore, assuming that each instrumental sound source is not always available in practice, we also investigate the effects of using instrumental sound source separation to obtain each source in the proposed method. Experimental results have shown that (1) unique similarity metrics can be learned for individual instrumental sound sources, (2) similarity metrics learned using some instrumental sound sources are possible to lead to more accurate results than that learned using the entire musical piece, (3) the performance degraded when learning with the separated instrumental sounds, and (4) similarity metrics learned by the proposed method well produced results that correspond to perception by human senses.
Deep neural network (DNN)-based speech enhancement usually uses a clean speech as a training target. However, it is hard to collect large amounts of clean speech because the recording is very costly. In other words, the performance of current speech enhancement has been limited by the amount of training data. To relax this limitation, Noisy-target Training (NyTT) that utilizes noisy speech as a training target has been proposed. Although it has been experimentally shown that NyTT can train a DNN without clean speech, a detailed analysis has not been conducted and its behavior has not been understood well. In this paper, we conduct various analyses to deepen our understanding of NyTT. In addition, based on the property of NyTT, we propose a refined method that is comparable to the method using clean speech. Furthermore, we show that we can improve the performance by using a huge amount of noisy speech with clean speech.
Research on automatic speech recognition (ASR) systems for electrolaryngeal speakers has been relatively unexplored due to small datasets. When training data is lacking in ASR, a large-scale pretraining and fine tuning framework is often sufficient to achieve high recognition rates; however, in electrolaryngeal speech, the domain shift between the pretraining and fine-tuning data is too large to overcome, limiting the maximum improvement of recognition rates. To resolve this, we propose an intermediate fine-tuning step that uses imperfect synthetic speech to close the domain shift gap between the pretraining and target data. Despite the imperfect synthetic data, we show the effectiveness of this on electrolaryngeal speech datasets, with improvements of 6.1% over the baseline that did not use imperfect synthetic speech. Results show how the intermediate fine-tuning stage focuses on learning the high-level inherent features of the imperfect synthetic data rather than the low-level features such as intelligibility.
Our previous work, the unified source-filter GAN (uSFGAN) vocoder, introduced a novel architecture based on the source-filter theory into the parallel waveform generative adversarial network to achieve high voice quality and pitch controllability. However, the high temporal resolution inputs result in high computation costs. Although the HiFi-GAN vocoder achieves fast high-fidelity voice generation thanks to the efficient upsampling-based generator architecture, the pitch controllability is severely limited. To realize a fast and pitch-controllable high-fidelity neural vocoder, we introduce the source-filter theory into HiFi-GAN by hierarchically conditioning the resonance filtering network on a well-estimated source excitation information. According to the experimental results, our proposed method outperforms HiFi-GAN and uSFGAN on a singing voice generation in voice quality and synthesis speed on a single CPU. Furthermore, unlike the uSFGAN vocoder, the proposed method can be easily adopted/integrated in real-time applications and end-to-end systems.
This paper describes the design of NNSVS, an open-source software for neural network-based singing voice synthesis research. NNSVS is inspired by Sinsy, an open-source pioneer in singing voice synthesis research, and provides many additional features such as multi-stream models, autoregressive fundamental frequency models, and neural vocoders. Furthermore, NNSVS provides extensive documentation and numerous scripts to build complete singing voice synthesis systems. Experimental results demonstrate that our best system significantly outperforms our reproduction of Sinsy and other baseline systems. The toolkit is available at https://github.com/nnsvs/nnsvs.
Sequence-to-sequence (seq2seq) voice conversion (VC) models have greater potential in converting electrolaryngeal (EL) speech to normal speech (EL2SP) compared to conventional VC models. However, EL2SP based on seq2seq VC requires a sufficiently large amount of parallel data for the model training and it suffers from significant performance degradation when the amount of training data is insufficient. To address this issue, we suggest a novel, two-stage strategy to optimize the performance on EL2SP based on seq2seq VC when a small amount of the parallel dataset is available. In contrast to utilizing high-quality data augmentations in previous studies, we first combine a large amount of imperfect synthetic parallel data of EL and normal speech, with the original dataset into VC training. Then, a second stage training is conducted with the original parallel dataset only. The results show that the proposed method progressively improves the performance of EL2SP based on seq2seq VC.
Neural-based text-to-speech (TTS) systems achieve very high-fidelity speech generation because of the rapid neural network developments. However, the huge labeled corpus and high computation cost requirements limit the possibility of developing a high-fidelity TTS system by small companies or individuals. On the other hand, a neural vocoder, which has been widely adopted for the speech generation in neural-based TTS systems, can be trained with a relatively small unlabeled corpus. Therefore, in this paper, we explore a general framework to develop a neural post-filter (NPF) for low-cost TTS systems using neural vocoders. A cyclical approach is proposed to tackle the acoustic and temporal mismatches (AM and TM) of developing an NPF. Both objective and subjective evaluations have been conducted to demonstrate the AM and TM problems and the effectiveness of the proposed framework.
We present a large-scale comparative study of self-supervised speech representation (S3R)-based voice conversion (VC). In the context of recognition-synthesis VC, S3Rs are attractive owing to their potential to replace expensive supervised representations such as phonetic posteriorgrams (PPGs), which are commonly adopted by state-of-the-art VC systems. Using S3PRL-VC, an open-source VC software we previously developed, we provide a series of in-depth objective and subjective analyses under three VC settings: intra-/cross-lingual any-to-one (A2O) and any-to-any (A2A) VC, using the voice conversion challenge 2020 (VCC2020) dataset. We investigated S3R-based VC in various aspects, including model type, multilinguality, and supervision. We also studied the effect of a post-discretization process with k-means clustering and showed how it improves in the A2A setting. Finally, the comparison with state-of-the-art VC systems demonstrates the competitiveness of S3R-based VC and also sheds light on the possible improving directions.
This paper presents a new voice conversion (VC) framework capable of dealing with both additive noise and reverberation, and its performance evaluation. There have been studied some VC researches focusing on real-world circumstances where speech data are interfered with background noise and reverberation. To deal with more practical conditions where no clean target dataset is available, one possible approach is zero-shot VC, but its performance tends to degrade compared with VC using sufficient amount of target speech data. To leverage large amount of noisy-reverberant target speech data, we propose a three-stage VC framework based on denoising process using a pretrained denoising model, dereverberation process using a dereverberation model, and VC process using a nonparallel VC model based on a variational autoencoder. The experimental results show that 1) noise and reverberation additively cause significant VC performance degradation, 2) the proposed method alleviates the adverse effects caused by both noise and reverberation, and significantly outperforms the baseline directly trained on the noisy-reverberant speech data, and 3) the potential degradation introduced by the denoising and dereverberation still causes noticeable adverse effects on VC performance.