Speech parameter generation algorithms for HMM-based speech synthesis

By Keiichi Tokuda, Takayoshi Yoshimura, Takashi Masuko, Takao Kobayashi, Tadashi Kitamura. 2000

1,177 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Speech & Audio

In Signal Processing

Citations in the corpus

Neural Speech Synthesis with Transformer Network (2018, methodology): “Parametric TTS systems (Tokuda et al. 2000; Zen, Tokuda, and Black 2009; Ze, Senior, and Schuster 2013; Tokuda et al. 2013) convert speech waves into spectrograms, and acoustic parameters, such as fundamental frequency and duration, are used to synthesize new audio results.”
A Survey on Neural Speech Synthesis (2021, methodology): “Compared to previous concatenative speech synthesis, HMMbased parametric synthesis is more flexible in changing speaker identities, emotions, and speaking styles [350].”
TTS-Portuguese Corpus: a corpus for speech synthesis in Brazilian Portuguese (2020, neutral): “In addition, Hidden Markov Model (HMM) based synthesis is still a topic of research (Aroon & Dhonde, 2015; Braude et al., 2013; Tokuda et al., 2000), as well as a variety of Unit Selection Models (Siddhi et al.”
Wavenet Based Low Rate Speech Coding (2017, background): “Ubiquitous are linear autoregressive (AR) models [7], and hidden Markov models (HMM) [8].”
Fast, Compact, and High Quality LSTM-RNN Based Statistical Parametric Speech Synthesizers for Mobile Devices (2016, methodology): “• Computation; With HMM-based SPSS, inference of acoustic parameters involves traversing decision trees at each HMM state and running the speech parameter generation algorithm [27].”
DelightfulTTS: The Microsoft Speech Synthesis System for Blizzard Challenge 2021 (2021, methodology): “Rich unit selection based concatenative speech synthesis [8] and statistical parametric speech synthesis like HMM/DNN/LSTM based models [9, 10, 11], have been the most popular methods in the past.”
Robust Speaker-Adaptive HMM-Based Text-to-Speech Synthesis (2009, methodology): “This research started by developing algorithms for generating a smooth parameter trajectory from HMMs [5]–[9].”
A Comparison of Recent Waveform Generation and Acoustic Modeling Methods for Neural-Network-Based Speech Synthesis (2018, neutral): “Neither formant enhancement nor MLPG [14] was used.”
The HMM-based speech synthesis system (HTS) version 2.0 (2007, methodology): “– Speech parameter generation algorithm (Case 1 in [15] only).”
Lightweight LPCNet-Based Neural Vocoder with Tensor Decomposition (2020, methodology): “At the time of testing, we applied MLPG [24] to predict smooth static features.”
Combining a vector space representation of linguistic context with a deep neural network for text-to-speech synthesis (2013, methodology): “After generating the acoustic features y k (including both static and dynamic features) from the DNN, we use the maximum likelihood parameter generation (MLPG) method [4] to generate smooth parameter trajectories to drive a vocoder.”
An Overview of Noise-Robust Automatic Speech Recognition (2014, methodology): “In [150], the pseudo-clean features generated with a HMM-based synthesis method [151] are used to replace the clean features which are usually hard to get in a real deployment.”
End-To-End Speech Synthesis Applied to Brazilian Portuguese (2020, methodology): “In addition, Hidden Markov Models (HMM) based synthesis are still under investigation [Tokuda et al., 2000], as well as a variety of Unit Selection Models [Wang and Georgila, 2011].”
Analysis of Speaker Adaptation Algorithms for HMM-Based Speech Synthesis and a Constrained SMAPLR Adaptation Algorithm (2009, positive): “A statistical parametric speech synthesis system based on hidden Markov models (HMMs) [6]–[12], in contrast, can easily and flexibly generate natural sounding synthetic speech with varying speaking styles and/or emotional expressions.”
Postfilters to Modify the Modulation Spectrum for Statistical Parametric Speech Synthesis (2016, methodology): “…generation algorithm in HMM-based TTS, 2) the proposed segment-level postﬁlter capable of achieving low-delay synthesis also yields signiﬁcant improvements in synthetic speech quality, and 3) the proposed postﬁlters are also effective in not only HMM-based TTS but also GMM-based VC and CLUSTERGEN.”
A New GAN-based End-to-End TTS Training Algorithm (2019, methodology): “From HMM [1] and DNN [2] to LSTM [3] and BLSTM [4], effective sequence modelling plays an important role in TTS.”