Speech recognition device, speech recognition method, and program

• US 7,813,928 B2
• Filed: 06/02/2005
• Issued: 10/12/2010
• Est. Priority Date: 06/10/2004
• Status: Expired due to Fees

• Current Assignee: Sovereign Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Inventors: Okimoto, Yoshiyuki, Inoue, Tsuyoshi, Tsuzuki, Takashi
• Primary Examiner(s): Han; Qi
• Application Number:
  US11/628,887
• Publication Number:
  US 20080167872A1
• Time in Patent Office: 1,958 Days
• Field of Search: 704/251, 704/231, 704/235, 704/236, 704/239, 704/255
• US Class Current: 704/251
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/221 Announcement of recognition results

Abstract

A speech recognition device presenting whether a user'"'"'s utterance is an unregistered word and whether the utterance should be repeated. The device includes a vocabulary storage unit (102) defining a vocabulary for speech recognition, and a speech recognition unit (101) checking the uttered speech against registered words. The device also includes a similarity calculation unit (103) calculating a similarity between the uttered speech and acoustic units, a judgment unit (104) judging, based on the check by the speech recognition unit (101) and the calculation performed by the similarity calculation unit (103), whether the uttered speech is a registered or unregistered word, an unregistered word unit (106) storing unregistered words, an unregistered word candidate search unit (105) searching the unregistered word unit (106) for unregistered word candidates, the, when the judgment unit (104) judges that the uttered speech is an unregistered word, and a display unit (107) displaying the result.

Claims

1. A speech recognition device that recognizes uttered speech and presents a result of the recognition of the uttered speech, the speech recognition device comprising:

a speech recognition word storage unit operable to define a vocabulary for speech recognition and operable to store the vocabulary for speech recognition as a plurality of registered words;

a speech recognition unit operable to check the uttered speech against the plurality of registered words stored in said speech recognition word storage unit;

a reference similarity calculation unit operable to search for a sequence having a highest similarity to the uttered speech, and operable to calculate a reference similarity regarding the sequence, the sequence being a combination of acoustic units, which are subwords;

an unregistered word judgment unit operable to judge whether the uttered speech is one of (i) a registered word of the plurality of registered words stored in said speech recognition word storage unit, and (ii) an unregistered word, which is not stored in said speech recognition word storage unit, said unregistered word judgment unit performing the judgment by comparing a similarity between the plurality of registered words checked against the uttered speech by said speech recognition unit and the uttered speech with the reference similarity calculated by said reference similarity calculation unit;

an unregistered word storage unit operable to store a plurality of unregistered words;

an unregistered word candidate search unit operable to search, from among the plurality unregistered words stored in said unregistered word storage unit, for an unregistered word candidate likely to correspond to the uttered speech, based on the uttered speech, said unregistered word candidate search unit performing the search when said unregistered word judgment unit judges that the uttered speech an unregistered word; and

a result display unit operable to display a result of the search performed by said unregistered word candidate search unit together with an indication that the uttered speech is an unregistered word.

2. The speech recognition device according to claim 1, wherein said unregistered word candidate search unit is operable to search for a plurality of unregistered word candidates from among the plurality of unregistered words stored in said unregistered word storage unit.

3. The speech recognition device according to claim 2, wherein said unregistered word storage unit is operable to store the plurality of unregistered words according to categories into which the plurality of unregistered words are classified.

4. The speech recognition device according to claim 1, wherein said unregistered word storage unit is operable to store the plurality of unregistered words according to categories into which the plurality of unregistered words are classified.

5. The speech recognition device according to claim 4, further comprising an unregistered word class judgment unit operable to judge a category of the unregistered word candidate based on the uttered speech.

6. The speech recognition device according to claim 5, wherein said unregistered word candidate search unit is operable to search for a plurality of unregistered word candidates from among the categories into which the plurality of unregistered words are classified in said unregistered word storage unit, said unregistered word candidate search unit performing the search for the plurality of unregistered word candidates based on a result of the judgment performed by said unregistered word class judgment unit.

7. The speech recognition device according to claim 4, further comprising an information acquisition unit operable to acquire information regarding the categories,

wherein said unregistered word candidate search unit is operable to search for a plurality of unregistered word candidates from among the categories into which the plurality of unregistered words are classified in said unregistered word storage unit, said unregistered word candidate search unit performing the search for the plurality of unregistered word candidates based on the information acquired by said information acquisition unit.

8. The speech recognition device according to claim 1, wherein said result display unit is operable to display the result of the search performed by said unregistered word candidate search unit, by excluding the plurality of registered words stored in said speech recognition word storage unit from the result of the search performed by said unregistered word candidate search unit.

9. The speech recognition device according to claim 1, wherein said unregistered word candidate search unit is operable to search for the unregistered word candidate by calculating an unregistered word score indicating a degree of similarity to the uttered speech.

10. The speech recognition device according to claim 9, wherein said result display unit is operable to display the unregistered word candidate and the calculated unregistered word score, as the result of the search performed by said unregistered word candidate search unit.

11. The speech recognition device according to claim 10, wherein said result display unit is operable to change the display of the unregistered word candidate according to the calculated unregistered word score.

12. The speech recognition device according to claim 1, wherein the plurality of unregistered words stored in said unregistered word storage unit are updated according to a predetermined condition.

13. The speech recognition device according to claim 12, further comprising a communication unit operable to communicate with an unregistered word server that stores unregistered words that are not stored in said unregistered word storage unit,

wherein the plurality of unregistered words stored in said unregistered word storage unit are updated when said communication unit receives the unregistered words stored by the unregistered word server.

14. The speech recognition device according to claim 1, wherein the plurality of registered words stored in said speech recognition word storage unit are updated according to a predetermined condition.

15. A speech recognition system for recognizing uttered speech and presenting a result of the recognition of the uttered speech, the speech recognition system comprising:

a speech recognition device that recognizes the uttered speech; and

an unregistered word search server that searches for an unregistered word not stored in said speech recognition device,

wherein said speech recognition device includes;

a speech recognition word storage unit operable to define a vocabulary for speech recognition and operable to store the vocabulary for speech recognition as a plurality of registered words;

a speech recognition unit operable to check the uttered speech against the plurality of registered words stored in said speech recognition word storage unit;

a reference similarity calculation unit operable to search for a sequence having a highest similarity to the uttered speech, and operable to calculate a reference similarity regarding the sequence, the sequence being a combination of acoustic units, which are subwords;

an unregistered word judgment unit operable to judge whether the uttered speech is one of (i) a registered word of the plurality of registered words stored in said speech recognition word storage unit, and (ii) an unregistered word not stored in said speech recognition word storage unit, said unregistered word judgment unit performing the judgment by comparing a similarity between the plurality of resisted words checked against the uttered speech by said speech recognition unit and the uttered speech with the reference similarity calculated by said reference similarity calculation unit;

a search request sending unit operable to send, to said unregistered word search server, a search request to search for an unregistered word candidate likely to correspond to the uttered speech, said search request sending unit sending the search request when said unregistered word judgment unit judges that the uttered speech is an unregistered word;

a search result receiving unit operable to acquire a result of the search for the unregistered word candidate from said unregistered word search server; and

a result display unit operable to display a result of the search for the unregistered word candidate together with an indication that the uttered speech is an unregistered word, and

wherein said unregistered word search server includes;

an unregistered word storage unit operable to store a plurality of unregistered words;

a search request receiving unit operable to receive the search request from said search request sending unit;

an unregistered word candidate search unit operable to search, from among the plurality of unregistered words stored in said unregistered word storage unit, for the unregistered word candidate likely to correspond to the uttered speech, based on the uttered speech, said unregistered word candidate search unit performing the search when said search request receiving unit receives the search request; and

a search result sending unit operable to send the result of the search performed by said unregistered word candidate search unit to said speech recognition device.

16. A speech recognition device that recognizes uttered speech and presents a result of the recognition of the uttered speech, in a speech recognition system including said speech recognition device and an unregistered word search server that searches for an unregistered word not stored in said speech recognition device, said speech recognition device comprising:

a speech recognition word storage unit operable to define a vocabulary for speech recognition and operable to store the vocabulary for speech recognition as a plurality of registered words;

a speech recognition unit operable to check the uttered speech against the plurality of registered words stored in said speech recognition word storage unit;

a reference similarity calculation unit operable to search for a sequence having a highest similarity to the uttered speech, and operable to calculate a reference similarity regarding the sequence, the sequence being a combination of acoustic units, which are subwords;

an unregistered word judgment unit operable to judge whether the uttered speech is one of (i) a registered word of the plurality of registered words stored in said speech recognition word storage unit, and (ii) an unregistered word not stored in said speech recognition word storage unit, said unregistered word judgment unit performing the judgment by comparing a similarity between the plurality of registered words checked against the uttered speech by said speech recognition unit and the uttered speech with the reference similarity calculated by said reference similarity calculation unit;

a search request sending unit operable to send, to the unregistered word search server, a search request to search for an unregistered word candidate likely to correspond to the uttered speech, said search request sending unit sending the search request when said unregistered word judgment unit judges that the uttered speech is an unregistered word;

a search result receiving unit operable to acquire a result of the search for the unregistered word candidate from the unregistered word search server; and

a result display unit operable to display the result of the search for the unregistered word candidate together with an indication that the uttered speech is an unregistered word.

17. A non-transitory computer-readable recording medium having a program recorded thereon, the program being executed by a speech recognition device that recognizes uttered speech and presents a result of the recognition of the uttered speech, the program causing a computer to execute a method comprising:

a speech recognition step of checking the uttered speech against a plurality of registered words stored in a speech recognition word database, which defines a vocabulary for speech recognition and stores the vocabulary for speech recognition as the plurality of registered words;

a reference similarity calculation step of searching for a sequence having a highest similarity to the uttered speech, and calculating a reference similarity regarding the sequence, the sequence being a combination of acoustic units, which are subwords;

an unregistered word judgment step of judging whether the uttered speech is one of (i) a registered word of the plurality of registered words stored in the speech recognition word database, and (ii) an unregistered word not stored in the speech recognition word database, said unregistered word judgment step performing the judgment by comparing a similarity between the plurality of registered words checked against the uttered speech in said speech recognition step and the uttered speech with the reference similarity calculated in said reference similarity calculation step;

an unregistered word candidate search step of searching, from among a plurality of unregistered words stored in an unregistered word database, for an unregistered word candidate likely to correspond to the uttered speech, based on the uttered speech, said unregistered word candidate search step performing the search when said unregistered word judgment step judges that the uttered speech is an unregistered word; and

a result display step of displaying a result of the search performed by said unregistered word candidate search step together with an indication that the uttered speech is an unregistered word.

18. A speech recognition method for recognizing uttered speech and presenting a result of the recognition of the uttered speech, the speech recognition method comprising:

a speech recognition step of checking the uttered speech against a plurality of registered words stored in a speech recognition word database which defines a vocabulary for speech recognition and stores the vocabulary for speech recognition as the plurality of registered words;

a reference similarity calculation step of searching for a sequence having a highest similarity to the uttered speech, and calculating a reference similarity regarding the sequence, the sequence being a combination of acoustic units, which are subwords;

an unregistered word judgment step of judging whether the uttered speech is one of (i) a registered word of the plurality of registered words stored in the speech recognition word database, and (ii) an unregistered word not stored in the speech recognition word database, said unregistered word judgment step performing the judgment by comparing a similarity between the plurality of registered words checked against the uttered speech in said speech recognition step and the uttered speech with the reference similarity calculated in said reference similarity calculation step;

an unregistered word candidate search step of searching, from among a plurality of unregistered words stored in an unregistered word database, for an unregistered word candidate likely to correspond to the uttered speech, based on the uttered speech, said unregistered word candidate search step performing the search when said unregistered word judgment step judges that the uttered speech is an unregistered word; and

a result display step of displaying a result of the search performed by said unregistered word candidate search step together with an indication that the uttered speech is an unregistered word.