Method and system for real time audio synthesis
AssigneeAMI Semiconductor, Inc.
InventorsHamid Sheikhzadeh-Nadjar, Etienne Cornu, Robert L. Brennan
Life of this patent
Estimating the term…
Abstract
A method and system for synthesizing audio speech is provided. A synthesis engine receives from a host, compressed and normalized speech units and prosodic information. The synthesis engine decompresses data and synthesizes audio signals. The synthesis engine can be implemented on a digital signal processing system which can meet requirements of low resources (i.e. low power consumption, lower memory usage), such as a DSP system including an input/output module, a WOLA filterbank and a DSP core that operate in parallel.
Claims50
A system for synthesizing audio signals that receives the text as input, analyses the text to find the speech unit labels and prosody parameters to provide speech units which are possibly compressed and prosody scripts which are possibly compressed, the system comprising:
- a decompression module for decompressing speech units and prosody scripts;
- and
- an overlap-add module for synthesizing speech using the speech units based on the prosody scripts.
A system for processing speech units, the system comprising:
- an offline compression module for compressing re-harmonized speech units;
- and
- an on-line frequency-domain decompression module having an oversampled synthesis filterbank for decompressing the compressed speech units.
A system for synthesizing audio signal, comprising:
- a decompression module for decompressing speech units, the speech unit including a frame of a constant pitch period;
- a circular shift pitch synchronous overlap-add (CS-PSOLA) module including a fixed-shift weighted overlap-add module for implementing a weighted overlap-add of the decompressed data, the circular shift pitch synchronous overlap-add module shifting the frame so that two consecutive frames make a periodic signal with a desired pitch period.
A system for synthesizing audio signals, the system comprising:
- an on-line processing module including;
- an interface for interfacing a host to receive compressed speech units and related compressed prosody parameters;
- a decompression module for decompressing data received on the interface;
- and
- an overlap-add module for synthesizing speech units using the speech units based on the related prosody parameters,
- the receipt of data from the host, decompression and speech synthesis are carried out in parallel, substantially in real-time.
A system for speech unit re-harmonization, the system comprising:
- an off-line module and an on-line module,
- the off-line module including;
- a normalizing module including a module for generating constant-pitch speech frames of more than one pitch period;
- a compression module for compressing the output of the normalizing module, and
- a database for recording the output of the compression module,
- the on-line module including;
- an interface for interfacing the off-line module for receiving data from the database;
- a decompression module for decompressing data received on the interface;
- and
- a speech engine for synthesizing speech using the output of the decompression module.
A method of synthesizing audio signals on a system that receives the text as input, analyses the text to find the speech unit labels and prosody parameters to provide speech units which are possibly compressed and prosody scripts which are possibly compressed, the method comprising the steps of:
- decompressing speech units and prosody scripts, and
- performing overlap-add synthesizing speech using the speech units based on the prosody scripts.
A method of synthesizing speech comprising the steps of:
- decompressing data regarding to speech units, the speech unit including at least one frame of a constant pitch period;
- and
- implementing a circular shift pitch synchronous overlap-add (CS-PSOLA), the CA-PSOLA step including the step of a fixed-shift weighted overlap-adding to applying a weighted, overlap-add process to the decompressed data, the step of the CS-PSOLA shifting the frame so that two consecutive frames make a periodic signal with a desired pitch period.
Dependencies are inferred from the claim text.