Speech recognition system

• US 8,566,091 B2
• Filed: 12/12/2007
• Issued: 10/22/2013
• Est. Priority Date: 12/12/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Knig, Lars, Hillebrecht, Christian, Lw, Andreas
• Primary Examiner(s): YEN, ERIC L
• Application Number:
  US11/955,233
• Publication Number:
  US 20090012792A1
• Time in Patent Office: 2,141 Days
• Field of Search: 704/236, 704/250, 704270-275
• US Class Current: 704/236
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A speech recognition system is provided for selecting, via a speech input, an item from a list of items. The speech recognition system detects a first speech input, recognizes the first speech input, compares the recognized first speech input with the list of items and generates a first candidate list of best matching items based on the comparison result. The system then informs the speaker of at least one of the best matching items of the first candidate list for a selection of an item by the speaker. If the intended item is not one of the best matching items presented to the speaker, the system then detects a second speech input, recognizes the second speech input, and generates a second candidate list of best matching items taking into account the comparison result obtained with the first speech input.

Claims

1. A computer-implemented speech recognition method for selecting, via a speech input, an intended item from a list of items, wherein the list of items includes items containing one or more words, the computer-implemented method comprising the following steps in association with a processor:

detecting a first speech input;

in a first computer process, recognizing the first speech input producing a first recognition result;

in a second computer process, comparing the first recognition result with the list of items and generating a first candidate list of best matching items based on the comparison result, the first candidate list containing a plurality of items and the first candidate list being a subset of the list of items;

presenting to the speaker at least one of the best matching items of the first candidate list for a selection of an item by the speaker, and if the intended item is not one of said best matching items presented to the speaker;

detecting a second speech input;

in a third computer process, recognizing the second speech input producing a second recognition result; and

in a fourth computer process, generating a second candidate list of best matching items matched to the second recognition result, the second candidate list being a subset of the first candidate list, and

in a fifth computer process, determining a score for each member of the second candidate list, each score being a measure of how well the second speech input corresponds to a respective member of the second candidate list, wherein when each of the determined scores is lower than a threshold value, the second speech input is compared to the complete list of items.

2. The speech recognition method of claim 1, where the second candidate list is generated by comparing the second recognition result to a sublist of the first candidate list.

3. The speech recognition method of claim 1, where the second candidate list is generated by subtracting the at least one of the best matching items of the first candidate list presented to the speaker from the first candidate list.

4. The speech recognition method of claim 3, where the second recognition result is compared to the second candidate list and a third candidate list of best matching items is generated based on the comparison result.

5. The speech recognition method of claim 4, where the N best results of the third candidate list are selected and presented to the speaker.

6. The speech recognition method of claim 4, further comprising comparing the second recognition result to the third candidate list and determining a score for each item of the third candidate list representing a measure of how well the second speech input fits to the items of the third candidate list.

7. The speech recognition method of claim 6, where, when the highest score of the third candidate list is lower than a predetermined threshold, the second speech input is compared to the complete list of items.

8. The speech recognition method of claim 6, further comprising:

receiving from the speaker an indication signal discarding the at least one of the best matching items of the first candidate list; and

removing the discarded items from the first candidate list;

where, when the highest score is lower than a predetermined threshold, the second recognition result is compared to the complete list of items with the items identified by the indication signal removed.

9. The speech recognition method of claim 1, where for recognizing the first and second speech input, a string of subword units for each speech input is generated, which is a recognition result, the string of subword units being compared with subword unit transcriptions of the list of items to identify matches, the matches resulting in the first and second candidate list, respectively.

10. The speech recognition method of claim 9, where the string of subword units for the second speech input is compared to the second candidate list resulting in a score, the score representing a measure of how well the string of subword units fits the subword unit transcriptions of the candidate list.

11. The speech recognition method of claim 4, where from the first and third candidate lists the best M results of the best matching items are displayed to the speaker for further selection, the speaker either selecting one of the displayed items or discarding all of the displayed items.

12. The speech recognition method of claim 1, where the speech input is used for selecting a destination location for a vehicle by a speech command or for selecting a media file out of a group of media files.

13. A speech recognition system for selecting, via a speech input, an intended item from a list of items, the system comprising:

a database containing a list of items stored in tangible non-transitory memory, the list of items having items containing one or more words;

a speech recognition module configured to recognize a first speech input resulting in a first recognition result, the speech recognition module including circuitry;

a comparing module configured to compare the first recognition result to the list of items and generating a first candidate list of best matching items based on the comparison result, the first candidate list containing a plurality of items and being a subset of the list of items;

an information unit configured to present to the speaker at least one of the best matching items of the first candidate list for a selection of an item by the speaker, and

a score determining unit, the system being configured so that if the intended item is not one of said best matching items presented to the speaker;

the speech recognition module recognizes a second speech input resulting in a second recognition result,

the comparing module generates a second candidate list of best matching items by comparing the second recognition result to the first candidate list, and

the score determining unit determines a score for each member of the second candidate list, each score being a measure of how well the second speech input corresponds to a respective member of the second candidate list, wherein when each of the determined scores is lower than a threshold value, the score comparison unit compares the second speech input to the list of items in the database.

14. The speech recognition system of claim 13, where the information unit comprises a display on which a subgroup from one of the first and second candidate lists is displayed.

15. The speech recognition system of claim 13, where the comparing module generates the second candidate list by subtracting the group of best matching items displayed, from the first candidate list.

16. The speech recognition system of claim 15, where the comparing module generates a third candidate list by comparing the second recognition result to the second candidate list.

17. The speech recognition system of claim 13 where the database contains subword unit transcriptions for each item.