Text-To-Speech Synthesis 5
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
A Phonologically Motivated Method of Selecting Non-Uniform Units
Authors:
Andrew P. Breen, BT Labs (U.K.)
Peter Jackson, BT Labs (U.K.)
Page (NA) Paper number 389
Abstract:
This paper describes a method for selecting units from a database of recorded speech, for use in a concatenative speech synthesiser. The simplest approach is to store one example of every possible unit. A more powerful method is to have multiple examples of each unit. The challenge for such a method is to provide an efficient means of selecting units from a practical inventory, to give the best approximation to the desired sequence in some clearly specified way. The method used in BT's Laureate system uses mixed N-phone units. In theory such units could be of arbitrary size, but in practice they are constrained to a maximum of three phones. It dynamically generates the unit sequence based on a global cost. Units are selected using purely phonologically motivated criteria, without reference to acoustic features, either desired or available within the inventory.
SL980389.PDF (From Author)
SL980389.PDF (Rasterized)
TOP
A Synthesis Method Based on Concatenation of Demisyllables and a Residual Excited Vocal Tract Model
Authors:
Steve Pearson, Panasonic Technologies, Inc./Speech Technology Lab (USA)
Nick Kibre, Panasonic Technologies, Inc./Speech Technology Lab (USA)
Nancy Niedzielski, Panasonic Technologies, Inc./Speech Technology Lab (USA)
Page (NA) Paper number 648
Abstract:
This paper describes the back-end of a new, flexible, high-quality TTS system. Preliminary results have demonstrated a highly natural and intelligible output. Although the system follows some standard methodologies, such as concatenation, we have introduced a number of novel features and a combination of techniques that make our system unique. We will describe in detail many of the design decisions and compare them with other known systems. A demonstration of the speech quality with implanted prosody is available in waveform file ([WAVE stltts1.wav and stltts2.wav]) on the conference CD.
SL980648.PDF (From Author)
SL980648.PDF (Rasterized)
0648_01.WAV(was: 0648_01.wav) | audio file demonstrating quality of synthesis method File type: Sound File Format: Sound File: WAV Tech. description: sample rate = 11025, mono, linear encoding, 16 bits per sample Creating Application:: Unknown Creating OS: Windows95 |
0648_02.WAV(was: 0648_02.wav) | audio file demonstrating quality of synthesis method File type: Sound File Format: Sound File: WAV Tech. description: sample rate = 11025, mono, linear encoding, 16 bits per sample Creating Application:: Unknown Creating OS: Windows95 |
TOP
Exploration of Acoustic Correlates in Speaker Selection for Concatenative Synthesis
Authors:
Ann K. Syrdal, AT&T Labs -- Research (USA)
Alistair Conkie, AT&T Labs -- Research (USA)
Yannis Stylianou, AT&T Labs -- Research (USA)
Page (NA) Paper number 882
Abstract:
It is often difficult to determine the suitability of a speaker to serve as a model for concatenative text-to-speech synthesis. The perceived quality of a speaker's natural voice is not necessarily predictive of its synthetic quality. The selection of female and male speakers on whom to base two synthetic voices for the new AT&T text-to-speech system was made empirically. Brief readings of identical text materials were recorded from professional speakers. Small-scale TTS systems were constructed with a minimal diphone inventory, suitable for synthesizing a limited number of test sentences. Synthesized sentences and their naturally spoken references were presented to listeners in a formal listening evaluation. In addition, a variety of acoustic measurements of the speakers were made in order to determine which acoustic characteristics correlated with subjective synthesis quality. The results have implications both for speaker selection and for improving concatenative synthesis methods.
SL980882.PDF (From Author)
SL980882.PDF (Rasterized)
TOP
A Perceptual Evaluation of Distance Measures for Concatenative Speech Synthesis
Authors:
Johan Wouters, Center for Spoken Language Understanding (USA)
Michael W. Macon, Center for Spoken Language Understanding (USA)
Page (NA) Paper number 905
Abstract:
In concatenative synthesis, new utterances are created by concatenating segments (units) of recorded speech. When the segments are extracted from a large speech corpus, a key issue is to select segments that will sound natural in a given phonetic context. Distance measures are often used for this task. However, little is known about the perceptual relevance of these measures. More insight into the relationship between computed distances and perceptual differences is needed to develop accurate unit selection algorithms, and to improve the quality of the resulting computer speech. In this paper, we develop a perceptual test to measure subtle phonetic differences between speech units. We use the perceptual data to evaluate several popular distance measures. The results show that distance measures that use frequency warping perform better than those that do not, and minimal extra advantage is gained by using weighted distances or delta features.
SL980905.PDF (From Author)
SL980905.PDF (Rasterized)
TOP
HMM-Based Smoothing For Concatenative Speech Synthesis
Authors:
Mike Plumpe, Microsoft Research (USA)
Alex Acero, Microsoft Research (USA)
Hsiao-Wuen Hon, Microsoft Research (USA)
Xuedong Huang, Microsoft Research (USA)
Page (NA) Paper number 908
Abstract:
This paper will focus on our recent efforts to further improve the acoustic quality of the Whistler Text-to-Speech engine. We have developed an advanced smoothing system that a small pilot study indicates significantly improves quality. We represent speech as being composed of a number of frames, where each frame can be synthesized from a parameter vector. Each frame is represented by a state in an HMM, where the output distribution of each state is a Gaussian random vector consisting of x and Dx. The set of vectors that maximizes the HMM probability is the representation of the smoothed speech output. This technique follows our traditional goal of developing methods whose parameters are automatically learned from data with minimal human intervention. The general framework is demonstrated to be robust by maintaining improved quality with a significant reduction in data.
SL980908.PDF (From Author)
SL980908.PDF (Rasterized)
TOP
A Nonlinear Unit Selection Strategy for Concatenative Speech Synthesis Based on Syllable Level Features
Authors:
Martin Holzapfel, SIEMENS AG (Germany)
Nick Campbell, ATR ITL (Japan)
Page (NA) Paper number 521
Abstract:
This paper describes an improved algorithm, motivated by fuzzy logic theory, for the selection of speech segments for concatenative synthesis from a huge database. Triphone HMM clustering is employed as an adaptive measure for articulatory similarity within a given database. Stress level contours are evaluated in the context of their surrounding vocalic peaks. The algorithm uses a beam search technique to optimise the suitability of each candidate unit to realise the desired target as well as continuity in concatenation.