Speech recognition apparatus and method using two opposite words

• US 6,937,982 B2
• Filed: 07/19/2001
• Issued: 08/30/2005
• Est. Priority Date: 07/21/2000
• Status: Active Grant

• Current Assignee: DENSO Corporation
• Original Assignee: DENSO Corporation
• Inventors: Ohno, Hiroshi, Kitaoka, Norihide
• Primary Examiner(s): McFadden, Susan
• Application Number:
  US09/907,594
• Publication Number:
  US 20020010579A1
• Time in Patent Office: 1,503 Days
• Field of Search: 704/231, 704/239, 704/243, 704/244, 704/251, 704/252, 704/255, 704/236, 704/237
• US Class Current: 704/252
• CPC Class Codes: G01C 21/3608 using speech input, e.g. using speech recognition
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

A speech recognition apparatus recognizes a speech signal received from a speaker and provides the result of recognition for an external device. In the apparatus, a pattern matching section performs pattern matching between each of reference patterns in a vocabulary and characteristic parameters extracted from the speech signal. The vocabulary includes reference patterns corresponding to words. Further the apparatus has a similar sound group which includes reference patterns corresponding to the sound similar to that of a specific word. The specific word is a word in response to which the external device performs an operation which cannot be easily undone. The speech signal is rerecognized by using the similar sound group. As a result, the pattern matching section outputs a word other than the specific word, if one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters.

Claims

1. A speech recognition apparatus for receiving and recognizing a speech signal from a speaker comprising:

an acoustic analysis means for analyzing a speech signal acoustically;

a feature extraction means for extracting characteristic parameters from the speech signal based on a result of the analysis performed by the acoustic analysis means; and

a pattern matching means for performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words, wherein;

the pattern matching means outputs as a result of recognition at least one word other than a specific word if the candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

the pattern matching means is connected to an external device, and the external device receives and uses the result of recognition from the pattern matching means for controlling an operation of the external device;

a similar sound group which includes reference patterns corresponding to sounds which are similar to but different from that of the specific word is stored beforehand;

the pattern matching means performs pattern matching between each of reference patterns in the similar sound group and the characteristic parameters if the candidate words include the specific word; and

the pattern matching means outputs as the result of recognition at least one word other than the specific word if one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters.

2. A speech recognition apparatus for receiving and recognizing a speech signal from a speaker comprising:

an acoustic analysis means for analyzing a speech signal acoustically;

a feature extraction means for extracting characteristic parameters from the speech signal based on a result of the analysis performed by the acoustic analysis means, and

a pattern matching means for performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words wherein;

the pattern matching means outputs as a result of recognition at least one word other than a specific word if the candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

the pattern matching means is connected to an external device, and the external device receives and uses the result of recognition from the pattern matching means for controlling an operation of the external device;

a similar sound group, which includes reference patterns corresponding to sounds that are similar to but different from that of the specific word, is stored beforehand;

the pattern matching means performs pattern matching between each of reference patterns in the similar sound ground and the characteristic parameters if the candidate words include the specific word;

the pattern matching means outputs as the result of recognition at least one word other than the specific word if one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters;

the similar sound group further includes reference patterns corresponding to sounds which are similar to but different from that of a second specific word that means the opposite to the specific word; and

the pattern matching means outputs as the result of recognition the second specific word if one of the reference pattern in the similar sound group has a high similarity with the characteristic parameters.

3. A speech recognition apparatus for receiving and recognizing a speech signal from a speaker comprising:

an acoustic analysis means for analyzing a speech signal acoustically;

a feature extraction means for extracting characteristic parameters from the speech signal based on a result of the analysis performed by the acoustic analysis means; and

a pattern matching means for performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words, wherein;

the pattern matching means outputs as a result of recognition at least one word other than a specific word if he candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

the pattern matching means is connected to an external device, and the external device receives and uses the result of recognition from the pattern matching means for controlling an operation of the external device; and

the pattern matching means outputs as a result of recognition at least one word other than the specific word if the candidate words include the specific word and an absolute level of confidence that the speech signal actually represents the specific word is low only in case that the speech signal is received in a situation where the speaker is prompted to reply to a query for confirming whether the speaker allows the external device to perform an operation which is not easily done.

4. A speech recognition apparatus for receiving and recognizing a speech signal from a speaker comprising:

a speech recognition means for recognizing the speech signal by using a vocabulary and outputting as a result of recognition at least one word in the vocabulary, the vocabulary being stored beforehand and including words;

a control means for receiving the result of recognition from the speech recognition means and outputting a control signal to an external device based on the result of recognition,

wherein the control means directs an output device to output a query to the speaker for confirming whether the speaker allows the external device to perform an operation if the control means receives as the result of recognition a word which directs the external device to perform the operation,

wherein the vocabulary includes a first word which allows the external device to perform an operation and a second word which inhibits the external device from performing an operation, and further includes similar words which are different from the first word but have acoustic characteristics similar to that of the first word, and

wherein the speech recognition means outputs the first word or the second word as a result of recognition of a reply to the query, and outputs the second word if the reply has a high similarity with one of the similar words.

5. A speech recognition apparatus as in claim 4, wherein the first word is an affirmative word and the second word is a negative word.

6. A speech recognition apparatus as in claim 5, wherein the external device is a navigation device.

7. A method for recognizing a speech signal comprising the steps of:

receiving a speech signal from a speaker;

analyzing the received speech signal acoustically;

extracting characteristic parameters from the speech signal based on a result of the analysis;

calculating similarities between each of reference patterns in a vocabulary and the extracted characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words;

selecting as candidate words at least one word corresponding to the reference pattern which has a high similarity with the characteristic parameters;

calculating similarities between each of reference patterns in a similar sound group and the characteristic parameters if the candidate words include a specific word, the similar sound group being stored beforehand and including the reference patterns corresponding to sounds that is similar to but different from that of the specific word;

outputting as a result of recognition at least one word other than the specific word if the candidate words include the specific word and one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters.

8. A method for recognizing a speech signal as in claim 7,

wherein the outputted result of recognition is received and used by an external device for controlling an operation of the device,

wherein the external device performs an operation which is not easily undone if it receives the specific word.

9. A speech recognition apparatus for receiving and recognizing speech signal from a speaker comprising:

an acoustic analysis means for analyzing the speech signal acoustically;

a feature extraction means for extracting characteristic parameters from the speech signal based on a result of the analysis performed by the acoustic analysis means; and

a pattern matching means for;

performing pattern matching between each of reference patterns in a previously memorized recognition object vocabulary and the extracted characteristic parameters; and

outputting as a recognition result a word that has a high matching (similarity) level, wherein;

the reference patterns in the recognition object vocabulary include reference patterns corresponding to a specific word and a group of reference patterns corresponding to an acoustic group similar to the specific word as a recognition object candidate group; and

when the pattern matching means performs the pattern matching, in a case that a reference pattern that has a high matching similarity level with the extracted characteristic parameters is included in the recognition object candidate group, the pattern matching means outputs a word different than the specific word.

10. The speech recognition apparatus of claim 9, wherein:

the recognition result by the pattern matching means is disposed in an external device for controlling an operation of the external device, and

the specific word is a word at has a possibility of adversely affecting the operation of the external device.

11. The speech recognition apparatus of claim 10, wherein the pattern matching means outputs the word different than the specific word when the speaker is prompted to execute a voice input to reply to a query for confirming whether the speaker allows the external device to perform a given operation.

12. The speech recognition apparatus of claim 10, wherein the external device whose operation is controlled by the recognition result by the pattern matching means is a navigation device.

13. A speech recognition method for receiving and recognizing speech signal from a speaker comprising:

analyzing the speech signal acoustically;

extracting characteristic parameters from the speech signal based on a result of the acoustic analysis; and

performing pattern matching between each of reference patterns in a previously memorized recognition object vocabulary and the extracted characteristic parameters; and

outputting as a recognition result a word that has a high matching (similarity) level, wherein;

the reference pattern in the recognition object vocabulary include reference patterns corresponding to a specific word and a group of reference patterns corresponding to an acoustic group similar the specific word as a recognition object candidate group; and

when the pattern matching is performed, in a case that a reference pattern that has a high matching similarity level with the extracted characteristic parameters is included in the recognition object candidate group, the pattern matching includes outputting a word different than the specific word.

14. The speech recognition method of claim 13, wherein:

the method includes using the recognition result in an external device for controlling an operation of the external device, and

the specific word is a word that has a possibility of adversely affecting the operation of the external device.

15. The speech recognition method of claim 14, wherein the pattern matching includes outputting the word different than the specific word when the speaker is prompted to execute a voice input to reply to a query for confirming whether the speaker allows the external device to perform a given operation.

16. The speech recognition apparatus of claim 14, wherein the external device whose operation is controlled by the recognition result of the pattern matching is a navigation device.

17. A speech recognition method for receiving and recognizing a speech signal from a speaker comprising:

analyzing a speech signal acoustically;

extracting characteristic parameters from the speech signal based on a result of the acoustic analysis; and

performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words, wherein;

the pattern matching outputs, as a result of recognition, at least one word other than a specific word if the candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

an external device receives and uses the result of recognition from the pattern matching for controlling an operation of the external device;

a similar sound group, which includes reference patterns corresponding to sounds that are similar to but different from that of the specific word, is stored beforehand;

the pattern matching includes pattern matching between each of reference patterns in the similar sound group and the characteristic parameters if the candidate words include the specific word; and

the pattern matching outputs, as the result of recognition, at least one word other than the specific word if one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters.

18. A speech recognition method for receiving and recognizing a speech signal from a speaker comprising:

analyzing a speech signal acoustically;

extracting characteristic parameters from the speech signal based on a result of the acoustic analysis; and

performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words, wherein;

the pattern matching outputs, as a result of recognition, at least one word other than a specific word if the candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

an external device receives and uses the result of recognition from the pattern matching for controlling an operation of the external device;

a similar sound ground, which includes reference patterns corresponding to sounds which are similar to but different from that of the specific word, is stored beforehand, and the pattern matching include pattern matching between each of reference patterns in the similar sound group and the characteristic parameters if the candidate words include the specific word;

the pattern matching includes outputting, as the result of recognition, at least one word other than the specific word if one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters;

the similar sound group further includes reference patterns corresponding to sounds that are similar to but different from that of a second specific word, which means the opposite to the specific word; and

the pattern matching includes outputting, as the result of recognition, the second specific word if one of the reference pattern in the similar sound group has a high similarity with the characteristic parameters.

19. A speech recognition method for receiving and recognizing a speech signal from a speaker comprising:

analyzing a speech signal acoustically;

extracting characteristic parameters from the speech signal based on a result of the acoustic analysis; and

performing pattern matching between each of reference patterns in a vocabulary and the extracted characteristic parameters and selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words, wherein;

the pattern matching includes outputting, as a result of recognition, at least one word other than a specific word if the candidate words include the specific word and a level of confidence that the speech signal actually represents the specific word is low;

an external device receives and uses the result of recognition from the pattern matching for controlling an operation of the external device; and

the pattern matching includes outputting, as a result of recognition, at least one word other than the specific word if the candidate words include the specific word and an absolute level of confidence that the speech signal actually represents the specific word is low only in case that the speech signal is received in a situation where the speaker is prompted to reply to a query for confirming whether the speaker allows the external device to perform an operation that is not easily done.

20. A speech recognition method for receiving and recognizing a speech signal from a speaker comprising:

recognizing the speech signal by using a vocabulary and outputting, as a result of recognition, at least one word in the vocabulary, the vocabulary being stored beforehand and including words;

receiving the result of recognition from the speech recognition and outputting a control signal to an external device based on the result of the recognition; and

outputting a query to the speaker for confirming whether the speaker allows the external device to perform operation if the received result of the recognition is a word that directs the external device to perform the operation, wherein;

the vocabulary includes a first word that allows the external device to perform an operation and a second word that inhibits the external device from performing an operation, and the vocabulary further includes similar words that are different from the first word but have acoustic characteristics similar to that of the first word; and

the recognizing includes outputting the first word or the second word as a result of recognition of a reply to the query and includes outputting the second word if the reply has a high similarity with one of the similar words.

21. A speech recognition method as in claim 20, wherein the first word is an affirmative word and the second word is a negative word.

22. A speech recognition method as in claim 20, wherein the external device is a navigation device.

23. An apparatus for recognizing a speech signal comprising:

means for receiving a speech signal from a speaker;

means for analyzing the received speech signal acoustically;

means for extracting characteristic parameters from the speech signal based on a result of the analysis;

means for calculating similarities between each of reference patterns in a vocabulary and the extracted characteristic parameters, the vocabulary being stored beforehand and including the reference patterns corresponding to words;

means for selecting as candidate words at least one word corresponding to the reference pattern that has a high similarity with the characteristic parameters;

means for calculating similarities between each of reference patterns in a similar sound group and the characteristic parameters if the candidate words include a specific word, the similar sound group being stored beforehand and including the reference patterns corresponding to sounds that is similar to but different from that of the specific word; and

means for outputting, as a result of recognition, at least one word other than the specific word if the candidate words include the specific word and one of the reference patterns in the similar sound group has a high similarity with the characteristic parameters.

24. An apparatus for recognizing a speech signal as in claim 23, wherein:

the outputted result of recognition is received and used by an external device for controlling an operation of the device; and

the external device performs an operation that is not easily undone if it receives the specific word.