Tonal correction of speech

• US 8,249,873 B2
• Filed: 08/12/2005
• Issued: 08/21/2012
• Est. Priority Date: 08/12/2005
• Status: Active Grant

• Current Assignee: Arlington Technologies, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Avaya Incorporated
• Inventors: Blair, Colin, Chan, Kevin, Gentle, Christopher R., Hepworth, Neil, Lang, Andrew W., Michaelis, Paul R.
• Primary Examiner(s): Godbold, Douglas
• Assistant Examiner(s): GUERRA-ERAZO, EDGAR X
• Application Number:
  US11/203,463
• Publication Number:
  US 20070038452A1
• Time in Patent Office: 2,566 Days
• Field of Search: 704/234, 704/255, 704/207, 704/257, 704/277, 704/270.1, 704/251, 704/260, 704/235, 704/243, 704/270, 704/276, 704/208, 704/4, 704/254, 704/231, 704/261, 381/56
• US Class Current: 704/252
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 25/15 the extracted parameters being formant information

Abstract

Tonal correction of speech is provided. Received speech is analyzed and compared to a table of commonly mispronounced phrases. These phrases are mapped to the phrase likely intended by the speaker. The phrase determines to be the phrase the user likely intended can be suggested to the user. If the user approves of the suggestion, tonal correction can be applied to the speech before that speech is delivered to a recipient.

Claims

1. A method for correction of recorded speech, comprising:

a communication device receiving a speech from a user, wherein the received speech includes at least two words;

a processor identifying at least a first received phrase in the received speech, wherein the identifying includes recognizing the at least two words of the first received phrase;

the processor comparing the first received phrase to a plurality of stored phrases, wherein the comparing includes scanning the plurality of stored phrases to match the first received phrase to a phrase included in the plurality of stored phrases, wherein the plurality stored phrases includes commonly mispronounced phrases, and wherein each of the commonly mispronounced phrases is mapped to at least one suggested phrase that includes at least two words, wherein the meanings of the commonly mispronounced phrases and of the at least one suggested phrase in each mapped pair is different;

the processor matching the first received phrase to at least a first one of the plurality of stored phrases comprising a commonly mispronounced phrase;

the processor determining from the matching that a second one of the plurality of stored phrases is likely an intended phrase based on likelihood of mispronunciation of the stored phrase, wherein the first one of the plurality of stored phrases is a commonly mispronounced phrase that is mapped to the second one of the plurality of stored phrases;

the processor suggesting by the communication device the second one of the plurality of stored phrases to the user as a first suggested substitute phrase, wherein a first word of the two words of the first received phrase is included in the first suggested substitute phrase, wherein a second word of the two words of the first received phrase is not included in the first suggested substitute phrase, wherein the first suggested substitute phrase is not provided to the user prior to the receiving speech from the user at the communication device; and

the processor automatically altering the first received phrase to correspond to the first suggested substitute phrase, wherein in automatically altering the first received phrase the processor corrects a tonal characteristic of at least a portion of the first received phrase.

2. The method of claim 1, further comprising:

receiving a confirmation signal from the user indicating that the second one of the plurality of stored phrases is the intended phrase;

in response to receiving the confirmation signal, altering the first received phrase to correspond to the second one of the plurality of stored phrases.

3. The method of claim 2, wherein the first received phrase is encoded using a vocal tract model comprising linear predictive coding.

4. The method of claim 2, wherein the altering the first received phrase comprises changing at least one of an amplitude and a pitch parameter of at least a portion of the first received phrase.

5. The method of claim 4, wherein spectral filtering parameters of the at least a portion of the first received phrase are not changed by the altering.

6. The method of claim 2, further comprising:

after the altering the first received phrase, releasing the first received phrase for delivery to a recipient mailbox.

7. The method of claim 2, wherein the altering the first received phrase includes changing a tone associated with at least one word of the first received phrase.

8. The method of claim 1, wherein the first received phrase and the second one of the plurality of stored phrases differ from one another in that a word of the first received phrase and a word of the second one of the plurality of stored phrases are associated with a different tone.

9. The method of claim 1, further comprising:

suggesting a number of the stored phrases to the user, wherein the number of stored phrases includes the second one of the plurality of stored phrases.

10. The method of claim 1, wherein the speech received from the user first received phrase is in a first language, and wherein the suggested second one of the plurality of stored phrases is presented to the user in a second language.

11. The method of claim 10, wherein the second one of the plurality of stored phrases is presented to the user by a text to speech application.

12. The method of claim 1, wherein a word included in the first received phrase is pronounced such that a tone associated with the pronunciation is indeterminate, and wherein the matching the first received phrase to at least a first one of the plurality of stored phrases comprises identifying at least a first one of the plurality of stored phrases that include the word associated with a first tone.

13. The method of claim 12, wherein the matching further comprises identifying a second one of the plurality of stored phrases that include the word associated with a second tone.

24. The method of claim 1, wherein the second word of the first received phrase differs only in tone from a second word of the first suggested substitute phrase, and wherein the first received phrase is non-sensical.

14. A system for the tonal correction of speech, comprising:

a user input, operable to receive speech;

a memory, wherein the memory contains a database of stored phrases, wherein the database of stored phrases includes commonly mispronounced phrases that are mapped to suggested substitute phrases;

a processor in communication with the user input and the memory, wherein in response to receipt of speech from a user from the user input, and wherein the speech comprises at least a first received phrase that includes at least two words, the processor operable to;

compare the first received phrase to at least a first stored phrase included in the database of stored phrases;

match the first received phrase to the first stored phrase, wherein the first stored phrase is mapped to a first suggested phrase; and

alter a first word in the first received phrase to match a second word in the first suggested phrase, wherein after altering the first word the first suggested phrase is identical to the first received phrase, wherein the alteration of a tone of the first word forms the second word that is different than the first word, wherein the first stored phrase is not provided to the user by the system prior to the receipt of speech by the system, and wherein the second word is output by the system.

15. The system of claim 14, wherein the first received phrase is digitized according to a vocal tract model, and wherein the alteration of a tone of the first word included in the first received phrase includes changing at least one of an amplitude and a pitch parameter of the first word.

16. The system of claim 14, further comprising:

a communication interface interconnected to the processor;

a communication network interconnected to the communication interface and to a plurality of addresses, wherein an altered first phrase is released for delivery to a recipient address.

17. The system of claim 16, further comprising:

a user output, wherein the first altered phrase has a meaning corresponding to the first suggested phrase, and wherein the first suggested phrase is presented to the user through the user output for approval, and wherein in response to confirmation from the user that the first suggested phrase is approved the first altered phrase is formed and is released for delivery to the recipient address.

18. The system of claim 17, wherein the user output includes a speaker, and wherein the first suggested phrase is presented to the user as speech.

19. The system of claim 16, further comprising:

a user output, wherein the first stored phrase is mapped to a number of suggested phrases, wherein the number of suggested phrases are presented to the user through the user output, wherein in response to user input comprising a selection of one of the suggested phrases the first altered phrase is formed, and wherein the first altered phrase corresponds to the selected one of the suggested phrases.

20. The system of claim 17, wherein the speech comprising at least a first received phrase is in a first language and wherein the first suggested phrase is presented to the user in a second language.

21. A system for correcting tonal speech, comprising:

means for receiving speech as input;

means for storing, wherein a plurality of stored phrases are stored in the means for storing, wherein the plurality of stored phrases include a plurality of commonly mispronounced phrases, wherein each commonly mispronounced phrase is mapped to a suggested substitute phrase, wherein each phrase includes at least two words;

means for comparing a first received phrase included in input speech to phrases included in the means for storing, wherein the first received phrase is matched to a first commonly mispronounced phrase included in the plurality of stored phrases, and wherein the first commonly mispronounced phrase is mapped to a first suggested substitute phrase;

means for altering a tone of at least a first word included in the first received phrase without further input by a user, wherein a meaning of the first received phrase is changed to correspond to a meaning of an associated phrase comprising the first suggested substitute phrase, wherein the first suggested substitute phrase is not provided to the user prior to receiving the first received phrase from the user, and wherein the first suggested substitute phrase is output.

22. The system of claim 21, further comprising:

means for outputting at least the first suggested substitute phrase to the user.

23. The system of claim 22, wherein altering a tone of at least at first word in the first received phrase creates an altered first received phrase, the system further comprising:

means for delivering the altered first received phrase to a recipient address in response to input from the user approving of the altered first received phrase.