Matsushita Electric Industrial Co., Ltd.

Steve D Pearson

Nicholas Kibre

Jean-Claude Junqua

Peter Veprek

Nancy Niedzielski

Roland Kuhn

Stephen L Johnson

Brian A Hanson

Frode Holm

Kazue Hata

Rhonda Shaw

Judson Hofmann

Ronald B Richard

The concatenative speech synthesizer employs demi-syllable subword units to generate speech. The synthesizer is based on a source-filter model that uses source signals that correspond closely to the human glottal source and that uses filter parameters that correspond closely to the human vocal tract. Concatenation of the demi-syllable units is facilitated by two separate cross face techniques, one applied in the time domain in the demi-syllable source signal waveforms, and one applied in the frequency domain by interpolating the corresponding filter parameters of the concatenated demi-syllables. The dual cross fade technique results in natural sounding synthesis that avoids time-domain glitches without degrading or smearing characteristic resonances in the filter domain.

Formant-based speech synthesizer employing demi-syllable concatenation with independent cross fade in the filter parameter and source domains

A speech synthesizer customization system provides a mechanism for generating a hierarchical customized user database. The customization system has a template management tool for generating the templates based on customization data from a user and associated replicated dynamic synthesis data from a text-to-speech (TTS) synthesizer. The replicated dynamic synthesis data is arranged in a dynamic data structure having hierarchical levels. The customization system further includes a user database that supplements a standard database of the synthesizer. The tool populates the user database with the templates such that the templates enable the user database to uniformly override subsequently generated speech synthesis data at all hierarchical levels of the dynamic data structure.

Method and tool for customization of speech synthesizer databases using hierarchical generalized speech templates

A speech synthesis system for generating voice dialog for a message frame having a fixed and a variable portion. A prosody module selects a prosodic template for each of the fixed and variable portions wherein at least one portion comprises a phrase of multiple words. An acoustic module selects an acoustic template for each of the fixed and variable portions wherein at least one portion comprises a phrase of multiple words. A frame generator concatenates the respective prosodic templates and acoustic templates. A sound module generates the voice dialog in accordance with the concatenated prosodic and acoustic templates.

Speech synthesis employing concatenated prosodic and acoustic templates for phrases of multiple words

An editing tool is provided for developing word-pronunciation pairs based on a spelled word input. The editing tool includes a transcription generator that receives the spelled word input from the user and generates a list of suggested phonetic transcriptions. The editor displays the list of suggested phonetic transcriptions to the user and provides a mechanism for selecting the desired pronunciation from the list of suggested phonetic transcriptions. The editing tool further includes a speech recognizer to aid the user in selecting the desired pronunciation from the list of suggested phonetic transcriptions based on speech data input that corresponds to the spelled word input, and a syllable editor that enables the user to manipulate a syllabic part of a selected pronunciation. Lastly, the desired pronunciation can be tested at any point through the use of a text-to-speech synthesizer that generates audible speech data for the selected phonetic transcription.

System for developing word-pronunciation pairs

Speech signal parameters are extracted from time-series data corresponding to different sound units containing the same vowel. The extracted parameters are used to train a statistical model, such as a Hidden Markov-based Model, that has a data structure for separately modeling the nuclear trajectory region of the vowel and its surrounding transition elements. The model is trained as through embedded re-estimation to automatically determine optimally aligned models that identify the nuclear trajectory region. The boundaries of the nuclear trajectory region serve to delimit the overlap region for subsequent sound unit concatenation.

Identification of unit overlap regions for concatenative speech synthesis system

An iterative formant analysis, based on minimizing the arc-length of various curves, and under various filter constraints estimates formant frequencies with desirable properties for text-to-speech applications. A class of arc-length cost functions may be employed. Some of these have analytic solutions and thus lend themselves well to applications requiring speed and reliability. The arc-length inverse filtering techniques are inherently pitch synchronous and are useful in realizing high quality pitch tracking and pitch epoch marking.

Extracting formant-based source-filter data for coding and synthesis employing cost function and inverse filtering

The concatenative speech synthesizer employs demi-syllable subword units to generate speech. The synthesizer is based on a source-filter model that uses source signals that correspond closely to the human glottal source and that uses filter parameters that correspond closely to the human vocal tract. Concatenation of the demi-syllable units is facilitated by two separate cross fade techniques, one applied in the time domain to the demi-syllable source signal waveforms, and one applied in the frequency domain by interpolating the corresponding filter parameters of the concatenated demi-syllables. The dual cross fade technique results in natural sounding synthesis that avoids time-domain glitches without degrading or smearing characteristic resonances in the filter domain.

Formant-based speech synthesizer employing demi-syllable concatenation

An electronic news receiving device receives text data for an electronic edition of a newspaper in the evening and audibly reads the newspaper to the user the next day. The newsreader includes a receiver which receives an electronic edition of a newspaper transmitted by a transmitter at an electronic news preparer's facility. The news preparer may mark words that do not conform to conventional letter-to-sound rules with a pronunciation flag and provide pronunciation data for the words with the electronic edition of the newspaper. The data may be compressed by providing a dictionary in the news receiving device and, at the preparer's facility, translating each word into a dictionary address. The transmission channel may include a telephone line, the vertical blanking interval of a television signal, a cable television channel, an AM/FM subcarrier signal or a satellite channel. The received electronic edition of the newspaper is processed by a section filter to retain desired sections of the newspaper and to discard unwanted sections. The retained news articles are stored in memory. A text-to-speech converter produces an audible output corresponding to the spoken text of the news articles. A user can input one or more keywords to cause the device to selectively read articles containing the keywords. The text to speech converter of the device uses rules and a dictionary to provide syntactic and semantic prosody for morpheme reconstructions. The user may determine which articles are read and may vary the rate at which articles are read using manual controls or spoken commands.

Electronic news reception apparatus that selectively retains sections

An application-independent, text-to-speech control system which employs an easy-to-use transport from which a user can control most text-to-speech conversion functions without prior training. It provides a control means by which user designated areas may be sequentially selected, stored and played. Furthermore, moving back and forward through user-selected text may be controlled by the user, with the synthesized speech output maintaining the proper pauses and inflections.

User interface controller for text-to-speech synthesizer

The voice source for the synthetic speech system is human generated speech waveforms that are inverse filtered to produce glottal waveforms representing larynx sound. These glottal waveforms are modified in pitch and amplitude, as required, to produce the desired sound. The human quality of the synthetically generated voice is further brought out by adding vocal tract effects, as desired. The pitch control is effected in one of two alternate ways, a loop method, or a concatenation method.

Growing community of inventors

Santa Barbara, CA, United States of America

Steve D Pearson