Voice to Text to Voice Processing

• US 20100324894A1
• Filed: 06/17/2009
• Published: 12/23/2010
• Est. Priority Date: 06/17/2009
• Status: Active Grant

• Current Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Original Assignee: Aristaeus Hermes LLC
• Inventors: Potkonjak, Miodrag
• Granted Patent: US 9,547,642 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G10L 13/00 Speech synthesis; Text to speech systems
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Technologies are generally described for voice to text to voice processing. An audio signal can be preprocessed and translated into text prior to being processed in the textual domain. The text domain processing or subsequent text to voice regeneration can seek to improve clarity, correct grammar, adjust vocabulary level, remove profanity, correct slang, alter dialect, alter accent, or provide other modifications of various oral communication characteristics. The processed text may be translated back into the audio domain for delivery to a listener. The processing at each stage may be driven by a set of objectives and constraints set by the speaker, the listener, a third party, or any combination of explicit or implicit participants. The voice processing may translate the voice content from a specific human language to the same human language with various improvements. The processing may also involve translation into one or more other languages.

Claims

1. A computer storage medium having computer-executable instructions stored thereon that configure the computer to:

receive a first voice audio signal

preprocess the first voice audio signal into a second voice audio signal

extract a first text language representation from the second voice audio signal;

transform the first text language representation into a second text language representation according to a set of language objectives;

transform the second text language representation into a third voice audio signal; and

provide the third voice audio signal as an update to the first voice audio signal.

2. The computer storage medium of claim 1, having computer-executable instructions stored thereon that further configure the computer to transform the first text language representation into a second text language representation according to a set of constraints.

3. The computer storage medium of claim 1, wherein the set of language objectives comprises a specification for correcting grammar, a specification for removing censored terms, a specification for improving grammatical style, a specification for creating acronyms, a specification for complexity adjustment, or a specification for vocabulary level adjustment.

4. The computer storage medium of claim 1, wherein the third voice audio signal comprises voice characteristics specified by a listener.

5. The computer storage medium of claim 1, wherein the set of language objectives comprises a specification for removing linguistic ambiguities, a specification for thesaurus-based word substitution, or a specification for expanding acronyms.

6. The computer storage medium of claim 1, having computer-executable instructions stored thereon that further configure the computer to transform the third voice audio signal by adding and additional audio signal.

7. A method for voice processing, the method comprising:

receiving a first voice audio signal;

transforming the first voice audio signal into a first text language representation of the first voice audio signal;

transforming the first text language representation into a second text language representation according to a set of language objectives and a set of constraints; and

sending the second text language representation to one or more components configured to regenerate a voice signal from the second text language representation.

8. The method of claim 7, further comprising transforming the regenerated voice signal by adding an additional audio signal.

9. The method of claim 7, wherein the set of language objectives comprises a specification for modifying sophistication level.

10. The method of claim 7, wherein transforming the first voice audio signal into a first text language representation comprises performing speech recognition operations trained to a set of voice characteristics associated with the first voice audio signal.

11. The method of claim 7, wherein the set of language objectives comprises a specification for removing ambiguities.

12. The method of claim 7, further comprising transmitting a representation of the first voice audio signal over a communication channel.

13. The method of claim 12, wherein the communication channel is a radio frequency channel.

14. A voice processing system comprising:

a processing unit;

a memory for storing an audio signal; and

a processing module configured to

receive a first voice audio signal,

extract a first text language representation from the first voice audio signal,

transform the first text language representation into a second text language representation according to a set of language objectives and a set of constraints, and

transform the second text language representation into a second voice audio signal.

15. The voice processing system of claim 14, wherein the set of language objectives comprises a specification for correcting grammar.

16. The voice processing system of claim 14, wherein the set of language objectives comprises a specification for modifying a linguistic complexity level.

17. The voice processing system of claim 14, wherein the processing module is further configured to transform the second voice audio signal by adding additional audio prior to providing the second voice audio signal to the speaker.

18. The voice processing system of claim 14, wherein transforming the first voice audio signal into a first text language representation comprises performing speech recognition operations trained to a set of voice characteristics associated with the first voice audio signal.

19. The voice processing system of claim 14, wherein the second voice audio signal comprises voice characteristics specified by a listener.

20. The voice processing system of claim 14, further comprising a storage medium, wherein the processing module is further configured to support time-shifting by storing a representation of the first voice audio signal onto the storage medium.