Code-Switching Speech Synthesis Based on Self-Supervised Learning and Domain Adaptive Speaker Encoder (original) (raw)
Related papers
A Comparison Of Deep Learning MOS Predictors For Speech Synthesis Quality
2023 34th Irish Signals and Systems Conference (ISSC)
This paper introduces a comparison of deep learning-based techniques for the MOS prediction task of synthesised speech in the Interspeech VoiceMOS challenge. Using the data from the main track of the VoiceMOS challenge we explore both existing predictors and propose new ones. We evaluate two groups of models: NISQA-based models and techniques based on fine-tuning the self-supervised learning (SSL) model wav2vec2 base. Our findings show that a simplified version of NISQA with 40% fewer parameters achieves results close to the original NISQA architecture on both utterance-level and system-level performances. Pre-training NISQA with the NISQA corpus improves utterance-level performance but shows no benefit on the system-level performance. Also, the NISQA-based models perform close to LD-Net and MOSANet, 2 out of 3 baselines of the challenge. Fine-tuning wav2vec2 base shows superior performance than the NISQA-based models. We explore the mismatch between natural and synthetic speech and discovered that the performance of the SSL model drops consistently when fine-tuned on natural speech samples. We show that adding CNN features with the SSL model does not improve the baseline performance. Finally, we show that the system type has an impact on the predictions of the non-SSL models.
Procedia Computer Science, 2021
Multi-speaker spoken datasets enable the creation of text-to-speech synthesis (TTS) systems which can output several voice identities. The multi-speaker (MSPK) scenario also enables the use of fewer training samples per speaker. However, in the resulting acoustic model, not all speakers exhibit the same synthetic quality, and some of the voice identities cannot be used at all. In this paper we evaluate the influence of the recording conditions, speaker gender, and speaker particularities over the quality of the synthesised output of a deep neural TTS architecture, namely Tacotron2. The evaluation is possible due to the use of a large Romanian parallel spoken corpus containing over 81 hours of data. Within this setup, we also evaluate the influence of different types of text representations: orthographic, phonetic, and phonetic extended with syllable boundaries and lexical stress markings. We evaluate the results of the MSPK system using the objective measures of equal error rate (EER) and word error rate (WER), and also look into the distances between natural and synthesised t-SNE projections of the embeddings computed by an accurate speaker verification network. The results show that there is indeed a large correlation between the recording conditions and the speaker's synthetic voice quality. The speaker gender does not influence the output, and that extending the input text representation with syllable boundaries and lexical stress information does not equally enhance the generated audio across all speaker identities. The visualisation of the t-SNE projections of the natural and synthesised speaker embeddings show that the acoustic model shifts some of the speakers' neural representation, but not all of them. As a result, these speakers have lower performances of the output speech.
Introducing Prosodic Speaker Identity for a Better Expressive Speech Synthesis Control
10th International Conference on Speech Prosody 2020, 2020
To have more control over Text-to-Speech (TTS) synthesis and to improve expressivity, it is necessary to disentangle prosodic information carried by the speaker's voice identity from the one belonging to linguistic properties. In this paper, we propose to analyze how information related to speaker voice identity affects a Deep Neural Network (DNN) based multi-speaker speech synthesis model. To do so, we feed the network with a vector encoding speaker information in addition to a set of basic linguistic features. We then compare three main speaker coding configurations: a) simple one-hot vector describing the speaker gender and identifier ; b) an embedding vector extracted from a speaker recognition pre-trained model ; c) a prosodic vector which summarizes information such as melody, intensity, and duration. To measure the impact of the input feature vector, we investigate the representation of the latent space at the output of the first layer of the network. The aim is to have an overview of our data representation and model behavior. Furthermore, we conducted a subjective assessment to validate the result. Results show that the prosodic identity of the speaker is captured by the model and therefore allows the user to control more precisely synthesis.
Autotuned voice cloning enabling multilingualism
This article describes a neural network-based text-to-speech (TTS) synthesis system that can generate spoken audio in a variety of speaker voices, including those not seen during training. We show that the proposed model can convert natural-language text-to-speech into a target language, and synthesize and translate natural text-to-speech. We quantify the importance of trained voice modules to obtain the best generalization performance. Finally, using randomly selected speaker embeddings, we show that speech can be synthesized with new speaker voices used in training and that the model learned high-quality speaker representations. We have also introduced a multilingual system and auto-tuner that allows you to translate regular text into another language, which makes multilingualization possible
Multi-speaker TTS with Deep Learning
2020
Recent advancements in technology have allowed for great development in the field of Speech Synthesis. As such, present-day speech synthesis applications are expected to function for multiple voices, and ensure a fast generation of natural–sounding synthetic speech for enhanced feasibility. This study suggests a multi-speaker text-to-speech (TTS) system for European Portuguese that enables the addition of new speakers without requiring extensive training and data. The proposed model framework comprises two systems: a sequence-to-sequence (Seq2Seq) regressive stage for acoustic feature prediction, followed by a neural vocoder for waveform generation. The model employs a universal vocoder which does not require fine-tuning for new voices. The Seq2Seq regressive model predicts acoustic features in the form of Mel-spectrograms by decoding the combination of linguistic embeddings — extracted from the text input —, and speaker embeddings conveying the target speaker identity. The model op...
Integrated speaker-adaptive speech synthesis
2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)
Enabling speech synthesis systems to rapidly adapt to sound like a particular speaker is an essential attribute for building personalised systems. For deep-learning based approaches, this is difficult as these networks use a highly distributed representation. It is not simple to interpret the model parameters, which complicates the adaptation process. To address this problem, speaker characteristics can be encapsulated in fixedlength speaker-specific Identity Vectors (iVectors), which are appended to the input of the synthesis network. Altering the iVector changes the nature of the synthesised speech. The challenge is to derive an optimal iVector for each speaker that encodes all the speaker attributes required for the synthesis system. The standard approach involves two separate stages: estimation of the iVectors for the training data; and training the synthesis network. This paper proposes an integrated training scheme for speaker adaptive speech synthesis. For the iVector extraction, an attention based mechanism, which is a function of the context labels, is used to combine the data from the target speaker. This attention mechanism, as well as nature of the features being merged, are optimised at the same time as the synthesis network parameters. This should yield an iVector-like speaker representation that is optimal for use with the synthesis system. The system is evaluated on the Voice Bank corpus. The resulting system automatically provides a sensible attention sequence and shows improved performance from the standard approach.
The Huya Multi-Speaker and Multi-Style Speech Synthesis System for M2voc Challenge 2020
ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2021
Text-to-speech systems now can generate speech that is hard to distinguish from human speech. In this paper, we propose the Huya multi-speaker and multi-style speech synthesis system which is based on DurIAN and HiFi-GAN to generate high-fidelity speech even under low-resource condition. We use the fine-grained linguistic representation which leverages the similarity in pronunciation between different languages and promotes the speech quality of code-switch speech synthesis. Our TTS system uses the HiFi-GAN as the neural vocoder which has higher synthesis stability for unseen speakers and can generate higher quality speech with noisy training data than WaveRNN in the challenge tasks. The model is trained on the datasets released by the organizer as well as CMU-ARCTIC, AIShell-1 and THCHS-30 as the external datasets and the results were evaluated by the organizer. We participated in all four tracks and three of them entered high score lists. The evaluation results show that our system outperforms the majority of all participating teams.
Learning Robust Latent Representations for Controllable Speech Synthesis
Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, 2021
State-of-the-art Variational Auto-Encoders (VAEs) for learning disentangled latent representations give impressive results in discovering features like pitch, pause duration, and accent in speech data, leading to highly controllable text-to-speech (TTS) synthesis. However, these LSTM-based VAEs fail to learn latent clusters of speaker attributes when trained on either limited or noisy datasets. Further, different latent variables start encoding the same features, limiting the control and expressiveness during speech synthesis. To resolve these issues, we propose RTI-VAE (Reordered Transformer with Information reduction VAE) where we minimize the mutual information between different latent variables and devise a modified Transformer architecture with layer reordering to learn controllable latent representations in speech data. We show that RTI-VAE reduces the cluster overlap of speaker attributes by at least 30% over LSTM-VAE and by at least 7% over vanilla Transformer-VAE.
Speaker verification-derived loss and data augmentation for DNN-based multispeaker speech synthesis
2021 29th European Signal Processing Conference (EUSIPCO), 2021
Building multispeaker neural network-based textto-speech synthesis systems commonly relies on the availability of large amounts of high quality recordings from each speaker and conditioning the training process on the speaker's identity or on a learned representation of it. However, when little data is available from each speaker, or the number of speakers is limited, the multispeaker TTS can be hard to train and will result in poor speaker similarity and naturalness. In order to address this issue, we explore two directions: forcing the network to learn a better speaker identity representation by appending an additional loss term; and augmenting the input data pertaining to each speaker using waveform manipulation methods. We show that both methods are efficient when evaluated with both objective and subjective measures. The additional loss term aids the speaker similarity, while the data augmentation improves the intelligibility of the multispeaker TTS system.
Continuous vocoder in feed-forward deep neural network based speech synthesis
2018
Recently in statistical parametric speech synthesis, we proposed a vocoder using continuous F0 in combination with Maximum Voiced Frequency (MVF), which was successfully used with hidden Markov model (HMM) based text-to-speech (TTS). However, HMMs often generate over-smoothed and muffled synthesized speech. From this point, we propose here to use the modified version of our continuous vocoder with deep neural networks (DNNs) for further improving its quality. Evaluations between DNNTTS using Continuous and WORLD vocoders are also presented. Experimental results from objective and subjective tests have shown that the DNN-TTS have higher naturalness than HMM-TTS, and the proposed framework provides quality similar to the WORLD vocoder, while being simpler in terms of the number of excitation parameters and models better the voiced/unvoiced speech regions than the WORLD vocoder.