Text to speech (TTS) has made rapid progress in both academia and industry in recent years. Some questions naturally arise that whether a TTS system can achieve human-level quality, how to define/judge that quality and how to achieve it. In this paper, we answer these questions by first defining the human-level quality based on the statistical significance of subjective measure and introducing appropriate guidelines to judge it, and then developing a TTS system called NaturalSpeech that achieves human-level quality on a benchmark dataset. Specifically, we leverage a variational autoencoder (VAE) for end-to-end text to waveform generation, with several key modules to enhance the capacity of the prior from text and reduce the complexity of the posterior from speech, including phoneme pre-training, differentiable duration modeling, bidirectional prior/posterior modeling, and a memory mechanism in VAE. Experiment evaluations on popular LJSpeech dataset show that our proposed NaturalSpeech achieves -0.01 CMOS (comparative mean opinion score) to human recordings at the sentence level, with Wilcoxon signed rank test at p-level p >> 0.05, which demonstrates no statistically significant difference from human recordings for the first time on this dataset.

While state-of-the-art Text-to-Speech systems can generate natural speech of very high quality at sentence level, they still meet great challenges in speech generation for paragraph / long-form reading. Such deficiencies are due to i) ignorance of cross-sentence contextual information, and ii) high computation and memory cost for long-form synthesis. To address these issues, this work develops a lightweight yet effective TTS system, ContextSpeech. Specifically, we first design a memory-cached recurrence mechanism to incorporate global text and speech context into sentence encoding. Then we construct hierarchically-structured textual semantics to broaden the scope for global context enhancement. Additionally, we integrate linearized self-attention to improve model efficiency. Experiments show that ContextSpeech significantly improves the voice quality and prosody expressiveness in paragraph reading with competitive model efficiency. Audio samples are available at: https://contextspeech.github.io/demo/

This article aims to improve neural TTS with vector-quantized, compact speech representations. We propose a Vector-Quantized Variational AutoEncoder (VQ-VAE) based feature analyzer to encode acoustic features into sequences with different time resolutions, and quantize them with multiple VQ codebooks to form the Multi-Stage Multi-Codebook Representation (MSMCR). The TTS system, MSMC-TTS, is proposed to predict better speech via this representation. In prediction, the multi-stage predictor is trained to map the input text sequence to MSMCRs in stages, by minimizing Euclidean distance and “triplet loss”. In synthesis, the neural vocoder converts ground-truth or predicted MSMCRs into speech waveforms. The proposed system is trained with single-speaker TTS datasets and tested in various scenarios for comprehensive evaluation. In TTS evaluation, MSMC-TTS obtains MOS of 4.34 and 4.10 on English and Chinese datasets, which significantly outperforms VITS with scores of 3.78 and 3.90. Meanwhile, compared with Mel-Spectrograms, the domain discrepancy between prediction and ground truth is lower in MSMCRs with the higher Domain-classification Error Rate (DER). Furthermore, this system shows lower modeling complexity and data size requirements, preserving excellent performance even with fewer model parameters or training data. The noticeable improvement in analysis-synthesis and TTS from multiple codebooks and stages also validate them as vital components in seeking a more profitable speech representation and building high-performance neural TTS.

The efficacy and robustness of Ordinal Regression (OR) in assessing speech pronunciation for language learning at phrase level has been shown before. However, for assessing phoneme pronunciation, we need to: 1. collect human scoring annotations for phoneme tokens of a short duration (60-70 ms); 2. train an ordinal regression model for each phoneme with the corresponding training and inference costs. In this paper, we propose to train a Universal Ordinal Regression (UOR) model instead of multiple, separate models for different phonemes, and evaluate its performance accordingly. A single universal binary classifier in UOR is trained to make a binary preference decision (better or worse) between a pair of two tokens with the same phoneme ID. In inference, labeled anchored tokens of specific phoneme ID in the training data are paired with test phoneme token to make binary preference decisions. By evaluating the new UOR on Speechocean762, a public speech database for pronunciation evaluation, we show the advantages of the proposed new approach. Improvements of Pearson Correlation Coefficient by 16.7% and Mean Square Error by 25.0%, all relatively, are obtained against the state-of-the-art systems.

End-to-end neural TTS has shown improved performance in speech style transfer. However, the improvement is still limited by the available training data in both target styles and speakers. Additionally, degenerated performance is observed when the trained TTS tries to transfer the speech to a target style from a new speaker with an unknown, arbitrary style. In this paper, we propose a new approach to seen and unseen style transfer training on disjoint, multi-style datasets, i.e., datasets of different styles are recorded, one individual style by one speaker in multiple utterances. An inverse autoregressive flow (IAF) technique is first introduced to improve the variational inference for learning an expressive style representation. A speaker encoder network is then developed for learning a discriminative speaker embedding, which is jointly trained with the rest neural TTS modules. The proposed approach of seen and unseen style transfer is effectively trained with six specifically-designed objectives: reconstruction loss, adversarial loss, style distortion loss, cycle consistency loss, style classification loss, and speaker classification loss. Experiments demonstrate, both objectively and subjectively, the effectiveness of the proposed approach for seen and unseen style transfer tasks. The performance of our approach is superior to and more robust than those of four other reference systems of prior art.

Ordinal regression with anchored reference samples (ORARS) has been proposed for predicting the subjective Mean Opinion Score (MOS) of input stimuli automatically. The ORARS addresses the MOS prediction problem by pairing a test sample with each of the pre-scored anchored reference samples. A trained binary classifier is then used to predict which sample, test or anchor, is better statistically. Posteriors of the binary preference decision are then used to predict the MOS of the test sample. In this paper, rigorous framework, analysis, and experiments to demonstrate that ORARS are advantageous over simple regressions are presented. The contributions of this work are: 1) Show that traditional regression can be reformulated into multiple preference tests to yield a better performance, which is confirmed with simulations experimentally; 2) Generalize ORARS to other regression problems and verify its effectiveness; 3) Provide some prerequisite conditions which can insure proper application of ORARS.

We propose a Multi-Stage, Multi-Codebook (MSMC) approach to high performance neural TTS synthesis. A vector-quantized, variational autoencoder (VQ-VAE) based feature analyzer is used to encode Mel spectrograms of speech training data by down-sampling progressively in multiple stages into MSMC Representations (MSMCRs) with different time resolutions, and quantizing them with multiple VQ codebooks, respectively. Multi-stage predictors are trained to map the input text sequence to MSMCRs progressively by minimizing a combined loss of the reconstruction Mean Square Error (MSE) and "triplet loss". In synthesis, the neural vocoder converts the predicted MSMCRs into final speech waveforms. The proposed approach is trained and tested with an English TTS database of 16 hours by a female speaker. The proposed TTS achieves an MOS score of 4.41, which outperforms the baseline with an MOS of 3.62. Compact versions of the proposed TTS with much less parameters can still preserve high MOS scores. Ablation studies show that both multiple stages and multiple codebooks are effective for achieving high TTS performance.

Recent advancements in neural end-to-end text-to-speech (TTS) models have shown high-quality, natural synthesized speech in a conventional sentence-based TTS. However, it is still challenging to reproduce similar high quality when a whole paragraph is considered in TTS, where a large amount of contextual information needs to be considered in building a paragraph-based TTS model. To alleviate the difficulty in training, we propose to model linguistic and prosodic information by considering cross-sentence, embedded structure in training. Three sub-modules, including linguistics-aware, prosody-aware and sentence-position networks, are trained together with a modified Tacotron2. Specifically, to learn the information embedded in a paragraph and the relations among the corresponding component sentences, we utilize linguistics-aware and prosody-aware networks. The information in a paragraph is captured by encoders and the inter-sentence information in a paragraph is learned with multi-head attention mechanisms. The relative sentence position in a paragraph is explicitly exploited by a sentence-position network. Trained on a storytelling audio-book corpus (4.08 hours), recorded by a female Mandarin Chinese speaker, the proposed TTS model demonstrates that it can produce rather natural and good-quality speech paragraph-wise. The cross-sentence contextual information, such as break and prosodic variations between consecutive sentences, can be better predicted and rendered than the sentence-based model. Tested on paragraph texts, of which the lengths are similar to, longer than, or much longer than the typical paragraph length of the training data, the TTS speech produced by the new model is consistently preferred over the sentence-based model in subjective tests and confirmed in objective measures.

Electroencephalography (EEG) is one of the most widely-used biosignal capturing technology for investigating brain activities, cognitive diseases, and affective disorders. To understand the underlying principles of brain activities and affective disorders using EEG data, one of the fundamental tasks is to accurately identify emotions from EEG signals, which has attracted huge attention in the field of affective computing. To improve the accuracy and effectiveness of emotion recognition based on EEG data, previous studies have successfully developed numerous feature extraction methods and classifiers. Among them, ensemble empirical mode decomposition (EEMD) is an efficient signal decomposition technique for extracting EEG features. It can alleviate the mode-mixing problem by adding white noise to the source signal. However, there remain some issues when applying this method to recognition tasks. As the added noise cannot be filtered completely, spurious modes are generated due to the residual noise. Therefore, it is crucial to perform intrinsic mode function (IMF) selection to find the most valuable IMF components that represent brain activities. Furthermore, the number of decomposed IMFs is various to different original signals, thus how to unify feature dimensions needs better solutions. To solve these issues, we propose a novel forecasting framework, named DEEMD-SPP, to identify emotions from EEG signals, based on the combination of denoising ensemble empirical mode decomposition (DEEMD) and Spatial Pyramid Pooling Network (SPP-Net). First, DEEMD is proposed to decompose the EEG signals, which effectively eliminates residual noise in the IMFs and selects the most valuable IMFs. Second, time-domain and frequency-domain features are extracted from the selected IMFs. Finally, SPP-net is employed as the classifier to recognize emotions, which can effectively transform various-sized feature maps into fixed-sized feature vectors through the pyramid pooling layer. The experimental results demonstrate that our proposed DEEMD-SPP framework can effectively reduce the effect of spike-in white noise, accurately extract EEG features, and significantly improve the performance of emotion recognition.

Many mispronunciation detection and diagnosis (MD&D) research approaches try to exploit both the acoustic and linguistic features as input. Yet the improvement of the performance is limited, partially due to the shortage of large amount annotated training data at the phoneme level. Phonetic embeddings, extracted from ASR models trained with huge amount of word level annotations, can serve as a good representation of the content of input speech, in a noise-robust and speaker-independent manner. These embeddings, when used as implicit phonetic supplementary information, can alleviate the data shortage of explicit phoneme annotations. We propose to utilize Acoustic, Phonetic and Linguistic (APL) embedding features jointly for building a more powerful MD&D system. Experimental results obtained on the L2-ARCTIC database show the proposed approach outperforms the baseline by 9.93%, 10.13% and 6.17% on the detection accuracy, diagnosis error rate and the F-measure, respectively.

FastSpeech, as a feed-forward transformer based TTS, can avoid the slow serial, autoregressive inference to generate the target mel-spectrogram in a parallel way. As a non-autoregressive TTS, the latency and computation load in inference is shifted from vocoder to transformer where the efficiency is limited by the quadratic time and memory complexity in the self-attention mechanism, particularly for a long text sequence. To tackle this challenges, We propose two models, ProbSparseFS and LinearizedFS, which have efficient self-attention arrangements to improve the inference speed and memory complexity. LinearizedFS has achieved 3.4x memory savings and 2.1x inference speedup, compared with the those of the baseline FastSpeech. A further optimized LinearizedFS with a lightweight FFN can accelerate the inference speed by 3.6x more. We do subjective voice quality evaluations in MOS and CMOS of News report and Audiobook applications, for multi-speaker and multi-style scenarios. Test results verified that the proposed models yield a TTS quality which is on-par with that of the baseline system but with much better memory efficiency and inference speed.

End-to-end TTS requires a large amount of speech/text paired data to cover all necessary knowledge, particularly how to pronounce different words in diverse contexts, so that a neural model may learn such knowledge accordingly. But in real applications, such high demand of training data is hard to be satisfied and additional knowledge often needs to be injected manually. For example, to capture pronunciation knowledge on languages without regular orthography, a complicated graphemeto-phoneme pipeline needs to be built based on a large structured pronunciation lexicon, leading to extra, sometimes high, costs to extend neural TTS to such languages. In this paper, we propose a framework to learn to automatically extract knowledge from unstructured external resources using a novel Token2Knowledge attention module. The framework is applied to build a TTS model named Neural Lexicon Reader that extracts pronunciations from raw lexicon texts in an end-to-end manner. Experiments show the proposed model significantly reduces pronunciation errors in low-resource, end-to-end Chinese TTS, and the lexicon-reading capability can be transferred to other languages with a smaller amount of data.

End-to-end neural TTS has achieved excellent performance on reading style speech synthesis. However, it is still a challenge to build a high-quality conversational TTS due to the limitations of corpus and modeling capability. This study aims at building a conversational TTS for a voice agent under sequence to sequence modeling framework. We firstly construct a spontaneous conversational speech corpus well designed for the voice agent with a new recording scheme ensuring both recording quality and conversational speaking style. Secondly, we propose a conversation context-aware end-to-end TTS approach that employs an auxiliary encoder and a conversational context encoder to specifically reinforce the information about the current utterance and its context in a conversation as well. Experimental results show that the proposed approach produces more natural prosody in accordance with the conversational context, with significant preference gains at both utterance-level and conversation-level. Moreover, we find that the model has the ability to express some spontaneous behaviors like fillers and repeated words, which makes the conversational speaking style more realistic.

End-to-end neural TTS training has shown improved performance in speech style transfer. However, the improvement is still limited by the training data in both target styles and speakers. Inadequate style transfer performance occurs when the trained TTS tries to transfer the speech to a target style from a new speaker with an unknown, arbitrary style. In this paper, we propose a new approach to style transfer for both seen and unseen styles, with disjoint, multi-style datasets, i.e., datasets of different styles are recorded, each individual style is by one speaker with multiple utterances. To encode the style information, we adopt an inverse autoregressive flow (IAF) structure to improve the variational inference. The whole system is optimized to minimize a weighed sum of four different loss functions: 1) a reconstruction loss to measure the distortions in both source and target reconstructions; 2) an adversarial loss to "fool" a well-trained discriminator; 3) a style distortion loss to measure the expected style loss after the transfer; 4) a cycle consistency loss to preserve the speaker identity of the source after the transfer. Experiments demonstrate, both objectively and subjectively, the effectiveness of the proposed approach for seen and unseen style transfer tasks. The performance of the new approach is better and more robust than those of four baseline systems of the prior art.

To scale neural speech synthesis to various real-world languages, we present a multilingual end-to-end framework that maps byte inputs to spectrograms, thus allowing arbitrary input scripts. Besides strong results on 40+ languages, the framework demonstrates capabilities to adapt to new languages under extreme low-resource and even few-shot scenarios of merely 40s transcribed recording, without the need of per-language resources like lexicon, extra corpus, auxiliary models, or linguistic expertise, thus ensuring scalability. While it retains satisfactory intelligibility and naturalness matching rich-resource models. Exhaustive comparative and ablation studies are performed to reveal the potential of the framework for low-resource languages. Furthermore, we propose a novel method to extract language-specific sub-networks in a multilingual model for a better understanding of its mechanism.

Recently, emotion classification from electroencephalogram (EEG) data has attracted much attention. As EEG is an unsteady and rapidly changing voltage signal, the features extracted from EEG usually change dramatically, whereas emotion states change gradually. Most existing feature extraction approaches do not consider these differences between EEG and emotion. Microstate analysis could capture important spatio-temporal properties of EEG signals. At the same time, it could reduce the fast-changing EEG signals to a sequence of prototypical topographical maps. While microstate analysis has been widely used to study brain function, few studies have used this method to analyze how brain responds to emotional auditory stimuli. In this study, the authors proposed a novel feature extraction method based on EEG microstates for emotion recognition. Determining the optimal number of microstates automatically is a challenge for applying microstate analysis to emotion. This research proposed dual-threshold-based atomize and agglomerate hierarchical clustering (DTAAHC) to determine the optimal number of microstate classes automatically. By using the proposed method to model the temporal dynamics of auditory emotion process, we extracted microstate characteristics as novel temporospatial features to improve the performance of emotion recognition from EEG signals. We evaluated the proposed method on two datasets. For public music-evoked EEG Dataset for Emotion Analysis using Physiological signals, the microstate analysis identified 10 microstates which together explained around 86% of the data in global field power peaks. The accuracy of emotion recognition achieved 75.8% in valence and 77.1% in arousal using microstate sequence characteristics as features. Compared to previous studies, the proposed method outperformed the current feature sets. For the speech-evoked EEG dataset, the microstate analysis identified nine microstates which together explained around 85% of the data. The accuracy of emotion recognition achieved 74.2% in valence and 72.3% in arousal using microstate sequence characteristics as features. The experimental results indicated that microstate characteristics can effectively improve the performance of emotion recognition from EEG signals.

Mean opinion score (MOS) is a popular subjective metric to assess the quality of synthesized speech, and usually involves multiple human judges to evaluate each speech utterance. To reduce the labor cost in MOS test, multiple methods have been proposed to automatically predict MOS scores. To our knowledge, for a speech utterance, all previous works only used the average of multiple scores from different judges as the training target and discarded the score of each individual judge, which did not well exploit the precious MOS training data. In this paper, we propose MBNet, a MOS predictor with a mean subnet and a bias subnet to better utilize every judge score in MOS datasets, where the mean subnet is used to predict the mean score of each utterance similar to that in previous works, and the bias subnet to predict the bias score (the difference between the mean score and each individual judge score) and capture the personal preference of individual judges. Experiments show that compared with MOSNet baseline that only leverages mean score for training, MBNet improves the system-level spearmans rank correlation coefficient (SRCC) by 2.9% on VCC 2018 dataset and 6.7% on VCC 2016 dataset.

This paper presents a speech BERT model to extract embedded prosody information in speech segments for improving the prosody of synthesized speech in neural text-to-speech (TTS). As a pre-trained model, it can learn prosody attributes from a large amount of speech data, which can utilize more data than the original training data used by the target TTS. The embedding is extracted from the previous segment of a fixed length in the proposed BERT. The extracted embedding is then used together with the mel-spectrogram to predict the following segment in the TTS decoder. Experimental results obtained by the Transformer TTS show that the proposed BERT can extract fine-grained, segment-level prosody, which is complementary to utterance-level prosody to improve the final prosody of the TTS speech. The objective distortions measured on a single speaker TTS are reduced between the generated speech and original recordings. Subjective listening tests also show that the proposed approach is favorably preferred over the TTS without the BERT prosody embedding module, for both in-domain and out-of-domain applications. For Microsoft professional, single/multiple speakers and the LJ Speaker in the public database, subjective preference is similarly confirmed with the new BERT prosody embedding. TTS demo audio samples are in https://judy44chen.github.io/TTSSpeechBERT/.

Text to speech (TTS), or speech synthesis, which aims to synthesize intelligible and natural speech given text, is a hot research topic in speech, language, and machine learning communities and has broad applications in the industry. As the development of deep learning and artificial intelligence, neural network-based TTS has significantly improved the quality of synthesized speech in recent years. In this paper, we conduct a comprehensive survey on neural TTS, aiming to provide a good understanding of current research and future trends. We focus on the key components in neural TTS, including text analysis, acoustic models, and vocoders, and several advanced topics, including fast TTS, low-resource TTS, robust TTS, expressive TTS, and adaptive TTS, etc. We further summarize resources related to TTS (e.g., datasets, opensource implementations) and discuss future research directions. This survey can serve both academic researchers and industry practitioners working on TTS.

A trajectory tiling based, hybrid TTS is revisited in this study for improving its synthesis performance. A combination of Transformer encoder and RNN based decoder architecture where two-level, at both word and Chinese phonetic alphabet letter levels, linguistic representation is exploited to generate a cogent and smooth speech parameter trajectory. And then a segment candidate lattice is constructed by minimizing the log spectral distortion of mel-spectrograms and RMSE of F0 between the generated trajectory and candidates. Normalized cross-correlation is used to find the best sequence of "wave-form tiles" in the lattice for synthesizing the final speech waveforms. Subjective A/B preference tests show that the new hybrid system outperforms our earlier trajectory-tiling hybrid baseline TTS (67% vs 11%) and the state-of-the-art, real-time TTS system constructed with Tacotron 2 and LPC-Net (56% vs 27%).