Method and system for generating natural sounding concatenative synthetic speech
View Patent ↗A method for generating synthetic speech can include identifying a recording of conversational speech and creating a transcription of the conversational speech. Using the transcription, rather than a predefined script, the recording can be analyzed and acoustic units extracted. Each acoustic unit can include a phoneme and/or a sub-phoneme. The acoustic units can be stored so that a concatenative text-to-speech engine can later splice the acoustic units together to produce synthetic speech.
1. A method for generating synthetic speech comprising the steps of;
presenting a stimulus to a human speaker;
prompting the human speaker to respond with conversational speech in response to the stimulus;
capturing a recording of the conversational speech spoken by the human speaker;
identifing a plurality of acoustic units from said recording, wherein each said acoustic unit includes at least one of a phoneme and a sub-phoneme;
extracting said acoustic units from said recording; and,
storing said acoustic units for use by a concatenative text-to-speech engine to generate synthetic speech, wherein the synthetic speech comprises one or more acoustic units spoken by the human speaker.
2. The method of claim 1 , further comprising the step of:
generating synthetic speech using said concatenative ten-to-speech engine using said acoustic units.
3. The method of claim 1 , further comprising the steps of:
determining prosodic information from said recording; and,
storing said prosodic information so that said prosodic information can be used by said text-to-speech engine when concatenating said acoustic units to form synthetic speech.
4. The method of claim 3 , further comprising the steps of:
generating a textual transcription from said recording, wherein said textual transcription is utilized in extracting said acoustic units, and wherein said textual transcription is utilized in determining said prosodic information.
5. The method of claim 1 , wherein said identifying step further comprises the steps of:
receiving conversational speech generated by a human speaker; and,
recording at least a portion of said conversational speech as said recording.
6. The method of claim 5 , wherein said receiving step further comprises the steps of:
establishing an acoustic environment;
positioning said human speaker within said acoustic environment, wherein the signal-to-noise ratio of said recorded conversational speech to other ambient noise recorded in said acoustic environment is at least 10 decibels; and,
prompting said human speaker to produce free form speech.
7. The method of claim 6 , wherein said prompting step further comprises the step of:
establishing a conversation between said human speaker and a second human speaker.
8. The method of claim 6 , wherein said prompting step further comprises prompting said human speaker using a prompting apparatus.
9. The method of claim 6 , wherein said signal-to-noise ratio is at least 30 decibels.
10. A system for synthetically generating speech comprising:
a training corpus containing at least one conversational speech recording that is captured by presenting a stimulus to a human speaker and prompting the human speaker to respond with conversational speech in response to the stimulus, and at least one associated transcription;
an acoustic unit store configured to store a plurality of acoustic units spoken by the human speaker, wherein at least a portion of said acoustic units are generated from data contained within said training corpus, and wherein at least a portion of said acoustic units are derived from said conversational speech recording;
a concatenative text-to-speech engine configured to utilize said acoustic unit store to synthetically generate speech, wherein said concatenative text-to-speech engine utilizes prosodic information extracted from said training corpus to synthetically generate speech; and
an acoustic environment within which conversational speech is recorded, wherein the signal-to-noise ratio of said recorded conversational speech to other ambient noise recorded in said acoustic environment is at least 10 decibels.
11. A machine-readable storage having stored thereon, a computer program having
a plurality of code sections, said code sections executable by a machine for causing the machine to perform the steps of:
presenting a stimulus to a human speaker;
prompting the human speaker to respond with conversational speech in response to the stimulus;
capturing a recording of the conversational speech spoken by the human speaker;
identifying a plurality of acoustic units from said recording, wherein each said acoustic unit includes at least one of a phoneme and a sub-phoneme;
extracting said acoustic units from said recording; and,
storing said acoustic units for use by a concatenative ten-to-speech engine to generate synthetic speech, wherein the synthetic speech comprises one or more acoustic units spoken by the human speaker.
12. The machine-readable storage of claim 11 , further comprising the step of:
generating synthetic speech using said concatenative text-to-speech engine using said acoustic units.
13. The machine-readable storage of claim 11 , further comprising the steps of:
determining prosodic information from said recording; and,
storing said prosodic information so that said prosodic information can be used by said text-to-speech engine when concatenating said acoustic units to form synthetic speech.
14. The machine-readable storage of claim 13 , further comprising the steps of:
generating a textual transcription from said recording, wherein said textual transcription is utilized in extracting said acoustic units, and wherein said textual transcription is utilized in determining said prosodic information.
15. The machine-readable storage of claim 11 , wherein said identifying step further comprises the steps of:
receiving conversational speech generated by a human speaker; and,
recording at least a portion of said conversational speech as said recording.
16. The machine-readable storage of claim 15 , wherein said receiving step further comprises the steps of:
establishing an acoustic environment;
disposing said human speaker within said acoustic environment, wherein the signal-to-noise ratio of said recorded conversational speech to other ambient noise recorded in said acoustic environment is at least 10 decibels; and,
prompting said human speaker to produce free form speech.
17. The machine-readable storage of claim 16 , wherein said prompting step further comprises the step of:
establishing a conversation between said human speaker and a second human speaker.
18. The machine-readable storage of claim 16 , wherein said prompting step further comprises prompting said human speaker using a prompting apparatus.
19. The machine-readable storage of claim 16 , wherein said signal-to-noise ratio is at least 30 decibels.