Text-To-Speech Synthesis 1
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Unsupervised Training of Phone Duration and Energy Models for Text-to-Speech Synthesis
Authors:
Paul C. Bagshaw, France Telecom, CNET. (France)
Page (NA) Paper number 132
Abstract:
A new model of phone duration and energy is presented. These parameters are modelled in two stages. The first stage builds a statistics tree that contains phone duration and energy mean and standard deviation values at each node. The branches of the tree are characterised by a set of factors related to phonetic context. The second stage considers phone duration and energy to be modified by two syllable-level prosodic coefficients. The duration and energy of the phones of a syllable are influenced to differing degrees by these coefficients. Weights are associated with the different phone positions in a syllable. A simulated annealing technique is used to find the set of weights that allow the prosodic coefficients to be calculated for all syllables and, in turn, minimise the error in predicting the phone duration and energy during synthesis. They are predicted with a mean squared error of 15.4ms and 6.8dB respectively. During synthesis, the syllable-level prosodic coefficients are predicted by regression trees from linguistic information. Manual prosodic labelling is not required at any stage.
SL980132.PDF (From Author)
SL980132.PDF (Rasterized)
TOP
Improved Duration Modeling of English Phonemes Using a Root Sinusoidal Transformation
Authors:
Jerome R. Bellegarda, Apple Computer, Inc. (USA)
Kim E.A. Silverman, Apple Computer, Inc. (USA)
Page (NA) Paper number 135
Abstract:
Accurate duration modeling is necessary for synthetic speech to sound natural. Over the past few years, the sums-of-products framework has emerged as an effective way to account for contextual influences on phoneme duration. This approach is generally applied after log-transforming the durations. This paper presents empirical and theoretical evidence which suggests that this transformation is not optimal. A promising alternative solution is proposed, based on a root sinusoidal function. Preliminary experimental results were obtained on over 50,000 phonemes in varied prosodic contexts. Compared to the log transformation, this new transformation reduced the proportion of standard deviation unexplained by approximately 30%. Alternatively, for a given level of performance, the root sinusoidal transformation roughly halved the number of regression parameters required.
SL980135.PDF (From Author)
SL980135.PDF (Scanned)
TOP
Efficient Adaptation of TTS Duration Model to New Speakers
Authors:
Chilin Shih, Bell Laboratories, Lucent Technologies (USA)
Wentao Gu, Shanghai Jiaotong University (China)
Jan P.H. van Santen, Bell Laboratories, Lucent Technologies (USA)
Page (NA) Paper number 177
Abstract:
This paper discusses a methodology using a minimal set of sentences to adapt an existing TTS duration model to capture inter-speaker variations. The assumption is that the original duration database contains information of both language-specific and speaker-specific duration characteristics. In training a duration model for a new speaker, only the speaker-specific information needs to be modeled, therefore the size of the training data can be reduced drastically. Results from several experiments are compared and discussed.
SL980177.PDF (From Author)
SL980177.PDF (Rasterized)
TOP
Duration Modeling For HMM-Based Speech Synthesis
Authors:
Takayoshi Yoshimura, Nagoya Institute of Technology (Japan)
Keiichi Tokuda, Nagoya Institute of Technology (Japan)
Takashi Masuko, Tokyo Institute of Technology (Japan)
Takao Kobayashi, Tokyo Institute of Technology (Japan)
Tadashi Kitamura, Nagoya Institute of Technology (Japan)
Page (NA) Paper number 939
Abstract:
This paper proposes a new approach to state duration modeling for HMM-based speech synthesis. A set of state durations of each phoneme HMM is modeled by a multi-dimensional Gaussian distribution, and duration models are clustered using a decision tree based context clustering technique. In the synthesis stage, state durations are determined by using the state duration models. In this paper, we take account of contextual factors such as stress-related factors and locational factors in addition to phone identity factors. Experimental results show that we can synthesize good quality speech with natural timing, and the speaking rate can be varied easily.